cis -Acting Element-specific Transcriptional Activity of Differentially Phosphorylated Nuclear Factor- (cid:1) B*

Phosphorylation of nuclear factor- (cid:1) B (NF- (cid:1) B) sub-units emerges as a mechanism by which transcriptional activity of nuclear NF- (cid:1) B complexes is regulated in an inhibitor (cid:1) B-independent fashion. As the main transactivator, the p65 subunit of NF- (cid:1) B has an outstanding position in the hierarchy of NF- (cid:1) B proteins. p65 is a multiply phosphorylated protein with phosphorylation sites in the C-terminal transactivation domain and the N-terminal Rel homology domain (RHD). In this study, we describe two previously non-reported phospho-ac-ceptor sites within the p65 RHD. We show that differential phosphorylation of serine residues within the RHD modulates transcriptional activity in a cis -acting element and promoter-specific context, thus leading to a phosphorylation state-dependent gene expression profile. RelA (cid:2) / (cid:2) mouse embryonic fibroblasts reconstituted with wild-type p65 or p65 phosphorylation-deficient mutants showed a distinctive expression profile of synthetic (cid:1) B-dependent reporters as well as endogenous genes. Hypophosphorylated p65 did not display cis -act-ing element-specific changes in DNA binding or dimerization behavior. This study shows for the first time that site-specific phosphorylation transiently transfecting the EcoPack2-293 ecotropic packaging cell supernatants 48 Viral titers were determined by titration on NIH 3T3 cells and RelA / MEFs were infected by exposing cells to retrovirus containing supernatants in the presence of 8 (cid:3) g/ml of Polybrene (Sigma) for 24 h. Stable transduced cell pools were selected in medium containing 300 (cid:3) g/ml Hygromycin B over 7 days, after cells were used for experiments. All experiments twice with cell pools derived from different infections. Gene Expression Analysis— Probes for Northern blot analysis were generated by polymerase chain reaction (PCR) from reverse transcribed RNA isolated from murine MEFs. Primers for generating a 709-bp fragment of the H2-K MHC class I gene product were 5 (cid:3) -CTG AAC GAA GAC CTG AAA ACG-3 (cid:3) and 5 (cid:3) -CTG TCA CCA AGT CCA CTC CAG-3 (cid:3) , respectively. A 489-bp fragment of the (cid:5) - actin gene was amplified using the primers 5 (cid:3) -ACC GTG AAA AGA TGA CCC AGA TC-3 (cid:3) and 5 (cid:3) -TAG

oligonucleotides coding for a tandem B site were cloned upstream of this minimal promoter using KpnI and SacI restriction sites. All oligonucleotides were designed to differ only in their decameric B consensus sequence. The prototypical sequence is 5Ј-GCT-decamer-CTG AGC TCC T-decamer-CTC AGC T-3Ј. The bicistronic retroviral vector pLXIH, which uses the 5Ј-long terminal repeat to drive expression of the transgene and the resistance gene was obtained by replacing the neomycin resistance gene of the pLXIN vector (Clontech, Palo Alto, CA) with a hygromycin resistance gene cassette. p65 wt and mutants were cloned into pLXIH by standard procedures. All constructs were verified by partial DNA sequencing using Dye Terminator chemistry. More detailed information on plasmids and cloning procedures can be obtained from the authors.
Reporter Gene Assays-Bovine aortic endothelial cells were transfected as described (20). RelA Ϫ/Ϫ MEFs were grown in 12-well plates and transfected at 70 -80% confluency. Cells were exposed to 400 ng of DNA (200 ng of reporter plasmid, 40 ng of p65 wt or mutant expression plasmid, 120 ng of pcDNA3, and 40 ng of cytomegalovirus/␤-galactosidase control plasmid) and 1.6 l of Lipofectamine (Invitrogen) in Dulbecco's modified Eagle's medium for 6 h. After addition of fetal bovine serum to a final concentration of 10%, cells were allowed to recover for 40 h. Cells were lysed with 0.075% Triton X-100 in 0.1 M KH 2 PO 4 , pH 7.8, and supernatants were assayed for luciferase and ␤-galactosidase activity as described (20).
Stable Transfectants-Stable RelA Ϫ/Ϫ MEFs expressing wt p65 or p65 mutants were generated by transfecting cells with pcDNA3 carrying wt p65 or mutants along with pcDNA3.1/Hygro (Invitrogen) in a 5:1 ratio. Single colonies were isolated by limiting dilution and selected in growth medium containing 300 g/ml Hygromycin B. Cell clones were screened by Western blot analysis for p65 expression. At least two different clones were used for experiments. Retrovirus containing supernatants were obtained after transiently transfecting the EcoPack2-293 ecotropic packaging cell line (Clontech) using Lipofectamine. Viruscontaining supernatants were harvested 48 and 72 h after transfection. Viral titers were determined by titration on NIH 3T3 cells and were Ͼ1 ϫ 10 5 /ml. RelA Ϫ/Ϫ MEFs were infected by exposing cells to retrovirus containing supernatants in the presence of 8 g/ml of Polybrene (Sigma) for 24 h. Stable transduced cell pools were selected in medium containing 300 g/ml Hygromycin B over 7 days, after which cells were used for experiments. All experiments were repeated twice with cell pools derived from different infections.
Gene Expression Analysis-Probes for Northern blot analysis were generated by polymerase chain reaction (PCR) from reverse transcribed RNA isolated from murine MEFs. Primers for generating a 709-bp fragment of the H2-K MHC class I gene product were 5Ј-CTG AAC GAA GAC CTG AAA ACG-3Ј and 5Ј-CTG TCA CCA AGT CCA CTC CAG-3Ј, respectively. A 489-bp fragment of the ␤-actin gene was amplified using the primers 5Ј-ACC GTG AAA AGA TGA CCC AGA TC-3Ј and 5Ј-TAG TTT CAT GGA TGC CAC AGG-3Ј. Northern blots were performed as described (20). Semi-quantitative RT-PCR was performed on singlestranded DNA that was generated with Superscript II reverse transcriptase and oligo-dT primers (Invitrogen) using 1 g of DNase Itreated total RNA as template. Reverse transcription was carried out according to the manufacturers' suggestions. Gene-specific primers were used to amplify fragments of the H2-K MHC class I (5Ј-CTG AAC GAA GAC CTG AAA ACG-3Ј and 5Ј-ATC AAC TCC TCC CCA TTC AAC-3Ј; 299-bp amplicon), vascular cell adhesion molecule-1 (VCAM-1; 5Ј-ACA CTC TTA CCT GTG CGC TGT-3Ј and 5Ј-ATT TCC CGG TAT CTT CAA TGG-3Ј; 314-bp amplicon), ICAM-1 (5Ј-TCC TAA AAT GAC  CTG CAG ACG-3Ј and 5Ј-AGT TTT ATG GCC TCC TCC TGA-3Ј; 314-bp  amplicon), and ␤-actin (for primers, see above) genes. The number of amplification cycles was chosen in the exponential range of the reaction and was 16 cycles for ␤-actin, 20 cycles for MHC class I, 30 cycles for VCAM-1, and 35 cycles for ICAM-1, respectively. Real-time quantitative PCR (qPCR) was carried out using SYBR Green chemistry (Invitrogen) on a Chromo4 continuous fluorescence monitoring thermocycler (MJ Research, Waltham, MA). The sequence of the primer set used in qPCR reactions to generate a 123-bp, cDNA-specific ICAM-1 amplicon was 5Ј-GCC TTG GTA GAG GTG ACT GAG-3Ј (forward) and 5Ј-GAC CGG AGC TGA AAA GTT GTA-3Ј (reverse). The sequence of the primer set to generate a 123-bp VCAM-1 amplicon was 5Ј-TGC CGA GCT AAA TTA CAC ATT G-3Ј (forward) and 5Ј-CCT TGT GGA GGG ATG TAC AGA-3Ј (reverse). The sequence of the primer set used in qPCR reactions to generate a 139-bp, cDNA-specific, interleukin-6 (IL-6) amplicon was 5Ј-ATG GAT GCT ACC AAA CTG GAT-3Ј (forward) and 5Ј-TGA AGG ACT CTG GCT TTG TCT-3Ј (reverse). An 81-bp amplicon for MHC class I was generated with primers 5Ј-GAT ACC TGA AGA ACG GGA ACG-3Ј (forward) and 5Ј-CTT CAG GTC TGC TGT GAT GG-3Ј (re-verse). A 94-bp amplicon specific for manganese superoxide dismutase (MnSOD) was generated with primers 5Ј-ACA GAT TGC TGC CTG  CTC TAA-3Ј (forward) and 5Ј-GTA GTA AGC GTG CTC CCA CAC-3Ј  (reverse). A 102-bp amplicon specific for monocyte inflammatory protein 2 (MIP-2) was generated with primers 5Ј-AAC ATC CAG AGC TTG  AGT GTG A-3Ј (forward) and 5Ј-TTC AGG GTC AAG GCA AAC TT-3Ј  (reverse). A primer set for the housekeeping gene hypoxanthine guanine  phosphoribosyl transferase was used to amplify a 103-bp amplicon from  total cDNA: forward, 5Ј-AGT GTT GGA TAC AGG CCA GAC-3Ј and  reverse, 5Ј-CGT GAT TCA AAT CCC TGA AGT-3Ј. Real-time PCR reactions were run in triplicates under the following conditions: a 10-min initial denaturing step at 94°C, 45 cycles of 15 s at 94°C, and 1 min at 60°C. At each cycle, fluorescence was quantified after product extension at 60°C. Product melting curves were generated for each reaction to ensure product fidelity. Quantitative changes in mRNA concentration were calculated according to Livak and Schmittgen (21).
Co-immunoprecipitations-RelA Ϫ/Ϫ MEFs were transiently transfected with expression vectors for wt p65 or p65 serine mutants together with a p50 expression plasmid. Twenty-four h after transfection, cells were treated with LPS (1 g/ml) for 45 min, washed with ice-cold PBS, and lysed for 30 min on ice in immunoprecipitation buffer (25 mM Hepes, pH 7.4, 10% glycerol, 150 mM NaCl, 0.1 mM EDTA, 0.5 mM dithiothreitol, and 0.5% Triton X-100) containing protease (Roche, Indianapolis, IN) and phosphatase (Sigma) inhibitor mixtures. Lysates were forced four times through a 26-gauge needle and cleared by centrifugation at 16,000 ϫ g for 10 min at 4°C. Equal amounts of lysate were incubated with 2 g of p50-specific polyclonal antibody (sc-114; Santa Cruz Biotechnology) together with 20 l of protein A-Sepharose slurry or with 20 l of agarose coupled anti-c-myc monoclonal antibody (sc-40AC, Santa Cruz Biotechnology) overnight at 4°C with continuous agitation. Immunocomplexes were washed four times in immunoprecipitation buffer, and precipitated proteins were eluted by boiling in Laemmli buffer. Proteins were separated by SDS-PAGE and transferred to polyvinylidene difluoride membranes. Membranes were probed with c-myc or p50-specific antibodies as indicated.

RESULTS
Identification and Characterization of p65 RHD Phosphoacceptor Sites-We used reporter gene assays together with a site-directed mutagenesis approach to identify potentially phosphorylated serine residues within p65 RHD. Eighteen single Ser-to-Ala mutants and one double mutant (p65 S238/240A) were expressed in bovine aortic endothelial cells and monitored for transcriptional activity using a NF-B-dependent reporter construct that carries three NF-B binding sites derived from the porcine E-selectin promoter (20). This approach revealed serine 205, 276, and 281 as being essential for p65 transcriptional activity (Fig. 1a). There was a limited effect of S42A and S75A substitutions on p65-mediated transcription, which was reduced by 50% in both mutants. The hypothesis that identified serines are indeed phosphorylated in vivo is strengthened by the fact that threonine substitutions at positions 205 and 276 were able to rescue p65 transcriptional activity (Fig. 1b), thus implying that these residues might be targeted by Ser/Thrdirected protein kinases. In contrast, Thr substitution at position 281 did not restore activity, which suggests that this position, if actually phosphorylated, is a substrate for a serinerestricted protein kinase. To address whether identified serines are phosphorylated in vivo, we prepared phospho-peptide maps of p65 proteins expressed in RelA Ϫ/Ϫ MEFs that have been exposed to LPS. All mutant proteins as well as wt p65 were multiply phosphorylated (Fig. 1c). Wild-type p65 showed at least six distinct phospho-peptides. All p65 mutants showed different phosphorylation patterns and lost at least one phos-pho-peptide species as compared with the wt protein. All mutants displayed residual phospho-peptides, indicating that p65 is multiply phosphorylated, as expected from previous results (19). It is noteworthy that the S205A mutant showed the largest decrease in phosphorylated peptide species. This finding could point to sequential phosphorylation, where a phospho-serine at position 205 would be required for other phosphorylation reactions to take place. Differences in intensities of separated phospho-peptides imply that not all serines are phosphorylated in the totality of cellular p65 proteins.
Differential p65 Phosphorylation Targets NF-B Activity to Gene Subsets-We first analyzed expression of three genes in RelA Ϫ/Ϫ MEFs derived clonal cell lines reconstituted with wt p65 or p65 phosphorylation deficient mutants. We determined LPS-induced mRNA levels for ICAM-1, VCAM-1, and MHC class I genes by Northern blotting and semi-quantitative RT-PCR, respectively. Although MHC class I mRNA was increased by wt p65 as well as all p65 mutants (Fig. 2, a and b), ICAM-1 was detected only in wt p65-expressing cells, although we observed, in two of five independent experiments, ICAM-1 mRNA in cells expressing the p65 S205A mutant. VCAM-1 was efficiently induced by wt p65 and to a lesser extent by p65 S205A. Analysis was extended using qPCR to analyze mRNA levels for ICAM-1, VCAM-1, MHC class I, IL-6, Mn-SOD, and MIP-2 gene products. For this set of experiments, FIG. 1. a, transcriptional activity of p65 Ser-to-Ala mutants on a NF-B-dependent promoter construct. Annotated positions of Ser-to-Ala substitutions correspond to the human p65 sequence. Fifty ng of wt p65 or mutant plasmids were transfected along with 700 ng of 3xB-Luciferase reporter construct and 100 ng of cytomegalovirus/␤-galactosidase into bovine aortic endothelial cells. Bars represent mean Ϯ S.E. (n ϭ 6, derived from three independent experiments) of galactosidase-normalized luciferase activity. wt p65-induced luciferase activity was set to 100. b, transcriptional activity of p65 Ser-to-Thr mutants. RelA Ϫ/Ϫ MEFs were transfected with p65 or mutant plasmid, along with a reporter construct harboring two NF-B sites derived from the porcine E-selectin promoter (n ϭ 6). The upper panel shows protein levels of wt p65 and respective mutants, as analyzed by Western blotting. c, two-dimensional phosphopeptide analysis of wt p65 and mutants expressed in RelA Ϫ/Ϫ MEFs. [ 32 P]Orthophosphate-labeled cells, which stably express wt p65 or mutants, were treated with LPS for 45 min. Tryptic digests of immunoprecipitated p65 were separated on cellulose plates and visualized by autoradiography. d, protein sequence alignment of different Rel proteins. p65, human, murine, rat; X. laevis, transforming protein (rel) homolog, Xenopus laevis; Dorsal, Drosophila melanogaster; v-rel, reticuloendotheliosis virus; c-Rel, human, murine, rabbit; RelB, human, murine; p105/p50, human, murine, chicken; p100/p52, human, murine. retrovirally transduced cell pools rather than single-cell clones were used. Furthermore, cells were activated by combined LPS (1 g/ml) and interferon-␥ (100 units/ml) treatment. All genes tested were induced after LPS/interferon-␥ exposure (Fig. 3). Induction in empty vector-transduced cells was limited (ICAM-1, 5-fold; VCAM-1, MHC class I, and MIP-2, 3-fold; IL-6, 70-fold; and MnSOD, 2-fold). In the resting state, wt p65-expressing cells showed higher mRNA levels of all genes analyzed than cells transduced with p65 serine mutants or empty vector. After LPS/interferon-␥ stimulation, expression profiles for ICAM-1, VCAM-1, and MHC class I genes were comparable with the ones obtained with LPS stimulation alone (Fig. 2). IL-6 was expressed in a p65 phosphorylation-dependent manner showing highest induction in wt p65 and lowest in p65 S281A-expressing cells. S205A and S276A mutants induced MnSOD as efficiently as wt p65, whereas the S281A mutants showed reduced expression levels. Similar to ICAM-1, the mouse GRO␣ analog MIP-2 was only efficiently induced by wt p65 and to a much lesser extent by the p65 S205A mutant.
The genomic organization of the respective enhancers is presented in Fig. 4. IL-6, MIP-2, and MnSOD intronic enhancer feature a single NF-B consensus site, whereas ICAM-1, VCAM-1, and MHC class I H2-K enhancers carry two B sites, one which is located far upstream (Ϫ1390) in the ICAM-1 promoter. Individual consensus sequences are highly conserved between mouse and human. In fact, only the B consensus in the MnSOD enhancer is changed at position ϩ3 (G for C) relative to the mouse sequence. Other cis-acting elements include the interferon responsive element in ICAM-1, VCAM-1, and MHC genes, Sp-1 consensus sites in ICAM-1, VCAM-1, and MIP-2 genes, CAAT/enhancer binding protein binding sites in ICAM-1, IL-6, and MnSOD, whereas IL-6 and MHC genes feature binding sites for members of the ATF/cyclic AMP-responsive element binding protein family of transcription factors.
Transcriptional Activity of Differentially Phosphorylated p65 Is cis-Acting Element Restricted-The most likely explanation for different p65 phosphorylation status-dependent gene transcription is that the sequence of a B consensus site within a promoter will dictate the requirement for p65 phosphorylation. To test this hypothesis, we investigated transcriptional activity of differentially phosphorylated p65 proteins on an array of decameric B consensus sequences. To evaluate different NF-B consensus sites, reporters were constructed in a way that only the decameric NF-B consensus site was altered. When reporters containing two copies of a given B consensus site were expressed alone or together with wt p65 or mutants, they could be divided into three groups according to their expression characteristics. Transcription from the first group (Fig. 5, a-e) of reporter constructs was only efficiently induced by wt p65. There was little or no induction when the reporters where co-transfected with p65 serine mutants. This group included highly asymmetric B sites that fit a GGRWWWYYYY consensus. Similarly, reporter constructs of the second group (Fig. 5, f-j) gave a characteristic expression profile. They were activated strongest by p65 wt, followed by the serine 205, 276, and 281 mutants with decreasing activity and are represented by a KGRAHWTYCC consensus. The third group consisted of binding sites that harbor four guanine bases in the 5Ј-half of the consensus sequence or represented a complete palindrome (Fig. 5, k-m). These constructs were induced by wt p65 and p65 mutants to a similar extent, with p65 S276A mutant showing the highest induction and which fit a GGGRATTYCC consensus. There was no transcriptional activity from a reporter construct harboring scrambled decamers, underlining that transcriptional activation is indeed achieved through the respective B binding sites and not through other sequences within the reporter construct (Fig. 5h). Group I showed a prevalence for a thymidine at the 3Ј-prime end (three of five) or a cytidine at position ϩ2 (three of five), both features not seen in group II and III consensus sequences. Binding sites within group II showed, with the exception of ICAM-1, an adherence to 5Ј-GGRA-3Ј at the 5Ј-end and to 5Ј-TYCC-3Ј at the 3Ј-end, while showing substantial heterogeneity at positions Ϫ1 and ϩ1. Group III sequences had a preference for four guanidines at the 5Ј-end (two of three) and featured a conservative TTYCC sequence in the 3Ј-half site. It is noteworthy that a single T-for-C substitution at the 3Ј-end can completely alter the transcriptional profile when engaged by differentially phosphorylated p65 proteins. Although transcription from the pig ELAM-1 B site (GGGAATTCCT) was only driven by wt p65, all mutant proteins where equally capable of driving transcription from a palindromic GGGAATTCCC consensus sequence.

DNA Binding and Dimerization Behavior of Differentially
Phosphorylated p65 Mutants-For DNA binding studies, we used one sequence of each group of B binding sites showing characteristic expression profiles when positioned upstream of a reporter gene. Thus, we studied NF-B binding to the IL-2 receptor-␣ B (GGGAATCTCC, group I), Ig light chain B (GGGACTTTCC, group II), and human ELAM-1 B (GGG-GATTTCC, group III) sites. Several DNA protein complexes with different electrophoretic mobility could be resolved (Fig. 6). The complex with the greatest mobility was composed of p50 homodimers as identified by specific removal of this complex by p50-specific antiserum (Fig 7a). This complex was found in all p65 and empty vector-transfected MEF extracts. RelA Ϫ/Ϫ MEFs transfected with wt p65 or p65 serine mutants but not empty vector-transfected cells showed two additional complexes that were identified as p50/p65 heterodimers and p65 homodimers by supershift analysis (Fig 7a). We could not detect any c-Relcontaining complexes. The relative abundance of p50 homodimers resulted most likely from the experimental design and limitations imposed by the transient transfection model, which results in 20 -30% transfected cells as assayed by flow cytometry using transfected green fluorescent protein as a marker. Because p50 homodimers are a major NF-B species in RelA Ϫ/Ϫ cells, this will result in a relative overpresentation of p50 homodimers in nuclear extracts. DNA-binding activity of p65 S281A containing complexes was lower than wt p65-or S205A-and S276Acontaining complexes. Interestingly, not all NF-B complexes seemed to have the same mobility, although we did not detect any difference in complex composition. p50/p65 heterodimers and p65 homodimers composed of the p65 S205A mutant were significantly retarded in comparison to wt p65-containing complexes. The retardation was also evident to a lesser extent in p65 S276Acontaining complexes. There was an additional complex in extracts derived from wt p65-expressing cells. This band was also visible to a lesser extent in extracts prepared from p65 S205Aexpressing MEFs. Because this band was shifted with p65specific antibody but not with p50-or c-Rel-specific antisera, we assume that the complex is a p65 degradation product also because it failed to shift with a p65 C-terminal antibody (data not shown). To test the possibility that p65 phosphorylation would alter dimerization behavior, we investigated association with p50 by co-immunoprecipitation. Cell extracts from wt p65-or p65 mutant-expressing cells were immunoprecipitated with p50-or c-myc-specific antiserum, which recognizes a short c-myc-derived tag sequence present in all p65 constructs. All p65 serine mutants co-immunoprecipitated with p50 as efficiently as the wt protein (Fig. 7b). Seemingly, p50 co-immunoprecipitated when extracts were incubated with c-myc tag recognizing antibody (Fig. 7c). Thus, p50/p65 dimerization is not regulated by p65 RHD phosphorylation. This is in agreement with the results obtained by EMSA, which did not reveal any differences in DNAbinding complex composition. DISCUSSION The transcription factor NF-B is an integrator of a vast array of inflammatory and immune-regulatory signals. The multiplicity of signals leading to NF-B activation is matched by a large set of genes that are dependent on its activity (6). It is unclear how specific signals are limiting their effect to a certain subset of genes. Specific gene activation by NF-B might be linked to the presence of different binding sites in the promoter region of respective genes, which results in a selective recruitment of NF-B dimers with higher affinity for these sites. However, affinity of certain NF-B dimers for a given cis-acting element should be stimulus independent and does not explain why a gene is not turned on by one stimulus while it is by another, although both are efficiently inducing IB degradation and NF-B nuclear translocation. Here we propose a model for transcriptional specificity, which is achieved through differential phosphorylation of a transactivator limiting its activity to a subset of genes. In addition to Ser-276 (15) and Ser-311 (17), we identified Ser-205 and Ser-281 as potential phospho-acceptor sites within p65 RHD. Comparative analysis of respective serines shows high conservation among species and within the protein family (Fig. 1d). Although there are Thr substitutions at positions 205 and 276, there are none at position 281, which is occupied by Ser in all Rel proteins, stressing the mandatory character of this residue. This is also highlighted by our finding that Thr is unable to substitute at this position, whereas it can at positions 205 and 276. This is in agreement with studies on the corresponding serine (Ser-342) of the p50 NF-B subunit (22). Similar to p65, p50 can be phosphorylated at the Ser-276 corresponding residue (Ser-337). In contrast to p65, however, p50 phosphorylation at Ser-337 and Ser-342 seem to be necessary for efficient DNA binding (22).
Transcriptional activity of hypophosphorylated p65 is not globally diminished but is rather dependent on the cis-acting element from which p65 exerts its activity. Although hypophosphorylated p65 can activate efficiently from one set of B elements, it has to be fully phosphorylated to achieve activation from another set. This might have physiological implications. It is possible that different extracellular signals lead to differentially phosphorylated p65, thereby restricting the cellular response to the expression of a specific subset of genes.
For instance in endothelial cells, H 2 O 2 , which induces NF-B nuclear translocation and DNA binding as efficiently as tumor necrosis factor-␣ (TNF-␣) fails to induce ICAM-1 expression. It was suggested that this might be linked to its failure to induce p65 phosphorylation (23). Furthermore, it has been shown that pharmacological inhibition of MSK-1 results in decreased p65 phosphorylation at Ser-276 and decreases IL-6 mRNA levels after TNF-␣ treatment. In contrast, TNF-␣ induction of NFkB2 mRNA was not inhibited by this treatment (16). The same study found that IL-6 was poorly induced by TNF-␣ in MSK1/ MSK2 Ϫ/Ϫ double knockout cells, whereas up-regulation of the NFkB2 gene was not affected. Similarly, we found that a reporter harboring an IL-6-derived B site (GGGATTTTCC) was only moderately induced by a p65 S276A mutant, whereas a reporter with palindromic B sites derived from the NFkB2 promoter (GGGAATTCCC) was induced independently of p65 phosphorylation status. Thus, it is possible that extracellular signals induce a selective subset of genes by altering the phosphorylation status of p65.
Other phosphorylation sites outside the RHD have been shown to be essential for p65 transcriptional activity. Ser-536, which is located in the C-terminal transactivation domain, is phosphorylated by inhibitor B kinases after cytokine stimu- lation and has been shown to be essential for p65 transcriptional activity (24). In addition, this residue might be targeted by ribosomal S6 kinase 1 (14) and the inhibitor B kinaserelated kinase T2K (25). At this point, it is not clear whether phosphorylation at S536 contributes to gene-specific transcription as seen with phospho-serines located within p65 RHD. We speculate that there may be no direct involvement of this residue in shaping cis-acting element-specific transcriptional activity, because it is unlikely that post-translational modifications within the C terminus will actively modulate RHD DNA interactions. Furthermore, we have shown that inhibition of RHD phosphorylation is diminishing p65 transcriptional activity, even in constructs were p65 TAD was replaced by VP16 TAD (19). Alternatively, C-terminal modifications may contribute to promoter-specific assembly of the basal transcriptional machinery and/or co-activators. That this happens in a cis-acting element-specific context is unlikely but has to be addressed in future studies.
This study shows that phosphorylation-deficient mutants bind to different cis-acting elements with similar affinities. Although there were some differences in DNA-binding activity of differentially phosphorylated p65 mutants, we did not detect changes in DNA-binding patterns that would account for differences in transcriptional activity seen in reporter assays. Only p65 S281A showed decreased DNA binding. This was, however, not specific for a certain consensus sequences and can therefore not explain differences in activity seen on different genes. This is consistent with previous findings that demonstrated that the p65 S276A mutant did not show altered DNAbinding activity (26) and that inhibition of p65 phosphorylation by a dominant-negative form of protein kinase C did not result in changed DNA affinity of NF-B complexes (19). On the other hand, it has been shown that in vitro phosphorylation of recombinant NF-B proteins enhances DNA binding (27). This finding is not in contradiction to our results, because single serine mutants used in our study are still multiply phosphorylated in vivo (Fig. 1c). Thus, other phosphorylation sites within RHD could compensate for the single phospho-serine loss caused by respective mutations, and the overall DNAbinding activity would remain unchanged. The altered electrophoretic mobility of some NF-B/DNA complexes can be attributed to three factors. First, differentially phosphorylated p65 species carry a different net charge that can alter electrophoretic mobility. In this context, it is notable that p65 S205A shows the biggest retardation, and at the same time it is the least phosphorylated p65 form, as assayed by phospho-peptide mapping (Fig. 1c). Second, binding of differentially phosphorylated p65 proteins to the consensus sequence could induce conformational changes in the DNA molecule, which would also result in different mobility of the protein-DNA complex. Changes in electrophoretic mobility attributable to alteration of DNA conformation upon NF-B binding have been reported (28). Third, phosphorylation can induce protein conformational changes that can affect electrophoretic separation.
The mechanism by which phosphorylation modulates p65 transcriptional activity in a cis-acting element-specific context remains elusive. Several models have to be considered. First, p65 phosphorylation could regulate its DNA-binding specificity. Although we did not see alterations in DNA binding that could explain the observed differences in transcriptional activity, there is still the possibility that in a cellular context there might be preferential recruitment of differentially phosphorylated p65 isoforms to a given promoter. Second, p65 phosphorylation could change the dimerization behavior of the protein. Therefore, phosphorylation might determine which other Rel protein is drawn into the dimeric NF-B complex. Different NF-B heterodimers have been shown to exhibit different transactivation potentials and favor different binding sites (29). However, we were unable to detect a phosphorylation-dependent change in NF-B dimer composition. All DNA-binding complexes could be removed with p50-or p65-specific antibodies in supershift assays, whereas c-Rel-specific antibodies had no effect. Furthermore, p50/p65 dimerization was independent of p65 phosphorylation status. Third, association with compo- nents of the transcriptional and/or chromatin modifying machinery could be regulated by phosphorylation. In this model, DNA binding and dimerization would not be altered, but interaction with co-activators, components of the polymerase II holoenzyme or histone-modifying proteins, could be modulated in a DNA-binding, site-specific context.
After submission of this report, Leung et al. (30) reported that a single nucleotide substitution within a B consensus site can alter the requirement for co-activators necessary to initiate NF-B-dependent transcription. It is speculated that a B consensus sequence induces specific conformational changes in the B dimer configuration and determines thereby which co-activator will interact with NF-B. Our results are consistent with that model and extend it to include an additional regulatory mechanism at the level of p65 phosphorylation. Thus, interaction of NF-B with certain co-activators might not only be dictated by DNA sequence but also by RHD phospho-serines. If B dimers bound to different consensus sites require different co-activators to initiate transcription, we speculate that the groups of consensus sites defined in our experiments will require different co-activators. Although the co-activator recruited to NF-B dimers bound to group I consensus sites can only be bound efficiently if p65 RHD is fully phosphorylated, a different co-activator that is recruited to dimers bound to group III consensus sites can bind to p65 in a phosphorylation-independent manner.
One additional possibility is that phosphorylation might change p65 acetylation status, as first hypothesized by Chen and Greene (31). Signal-induced acetylation has emerged as an important mechanism to regulate p65 subcellular localization, interaction with inhibitor Bs, DNA binding, and transcriptional activity (32,33). It is thus possible that site-specific phosphorylation dictates the acetylation pattern of the protein and thereby regulates its function.
In summary, we propose a p65 "phosphorylation code" that targets NF-B activity to distinct subsets of genes. Cells and organs are continuously exposed to a vast variety of extracellular signals, which have to be integrated to achieve signalspecific transcriptional programs. In this study we show, with the example of NF-B, that transcriptional selectivity can be achieved not only by activating different signaling cascades leading to the engagement of distinct transcriptional activators or repressors but also at the level of the transcription factor itself, which is targeted to specific gene subsets by altering its phosphorylation status. Future studies will show whether this regulatory system applies to other transcription factors known to be regulated by phosphorylation.