Direct phosphorylation of NF-kappaB1 p105 by the IkappaB kinase complex on serine 927 is essential for signal-induced p105 proteolysis.

The p105 precursor protein of NF-kappaB1 acts as an NF-kappaB inhibitory protein, retaining associated Rel subunits in the cytoplasm of unstimulated cells. Tumor necrosis factor alpha (TNFalpha) and interleukin-1alpha (IL-1alpha) stimulate p105 degradation, releasing associated Rel subunits to translocate into the nucleus. By using knockout embryonic fibroblasts, it was first established that the IkappaB kinase (IKK) complex is essential for these pro-inflammatory cytokines to trigger efficiently p105 degradation. The p105 PEST domain contains a motif (Asp-Ser(927)-Gly-Val-Glu-Thr), related to the IKK target sequence in IkappaBalpha, which is conserved between human, mouse, rat, and chicken p105. Analysis of a panel of human p105 mutants in which serine/threonine residues within and adjacent to this motif were individually changed to alanine established that only serine 927 is essential for p105 proteolysis triggered by IKK2 overexpression. This residue is also required for TNFalpha and IL-1alpha to stimulate p105 degradation. By using a specific anti-phosphopeptide antibody, it was confirmed that IKK2 overexpression induces serine 927 phosphorylation of co-transfected p105 and that endogenous p105 is also rapidly phosphorylated on this residue after TNFalpha or IL-1alpha stimulation. In vitro kinase assays with purified proteins demonstrated that both IKK1 and IKK2 can directly phosphorylate p105 on serine 927. Together these experiments indicate that the IKK complex regulates the signal-induced proteolysis of NF-kappaB1 p105 by direct phosphorylation of serine 927 in its PEST domain.

In unstimulated cells, NF-B is sequestered in the cytoplasm in an inactive form bound to inhibitory proteins, termed IBs. In response to agonist stimulation with ligands such as the pro-inflammatory cytokines tumor necrosis factor (TNF) ␣ and interleukin-1 (IL-1), IBs are proteolytically degraded, releasing associated NF-B dimers to translocate into the nucleus and induce gene expression (1,2). IBs consist of a family of structurally related proteins, which includes IB␣, IB␤, and IB⑀ together with the precursor forms of NF-B1 (p105) and NF-B2 (p100) (3). These all contain multiple ankyrin repeats that interact with the nuclear localizing signals of RHDs to prevent nuclear translocation of associated Rel subunits.
The best characterized member of the IB family is IB␣, which binds p50/Rel-A and p50/c-Rel heterodimers (3). In response to stimulation with an NF-B agonist, IB␣ is rapidly phosphorylated in a conserved D(P)SGXS(P) motif at its N terminus (8 -11). The regulatory serines of IB␣ are phosphorylated by a 700-kDa complex, the IB kinase (IKK) complex, which contains three polypeptides as follows: the serine kinases IKK1 (IKK␣) and IKK2 (IKK␤) and a structural subunit NEMO (IKK␥) (2). IKK␣ and IKK␤ form a heterodimer that directly phosphorylates IB␣ (12) and is activated by both TNF␣ and IL-1. NEMO plays an essential role in coupling the IKK complex to upstream signals (13). The phosphorylated sequence is recognized by the SCF ␤TRCP ubiquitin ligase complex which then promotes rapid ubiquitination on two adjacent lysine residues and subsequent degradation of IB␣ by the 26 S proteasome to release associated NF-B dimers (14,15).
The constitutive proteolytic generation of NF-B1 p50 from its precursor p105 involves ubiquitination and is mediated by the 26 S proteasome (6). This is an unusual example of limited proteolysis by the proteasome resulting from the presence of a glycine-rich region in the C-terminal half of the p50 moiety of p105 that appears to act as a physical barrier to proteasome entry (16,17). Processing to p50 is inefficient, and the majority of p105 is simply slowly degraded. Therefore, p105 levels may be regulated by two proteolytic pathways, limited (processing to p50) and complete (degradation). An alternative mechanism has also been suggested in which p105 is processed co-translationally by the proteasome to produce p50 (18). The relative importance of post-translational versus co-translational processing of p105 is presently unclear and may be cell type-specific.
Unprocessed p105 functions as an IB through the association of its C-terminal ankyrin repeats with p50, c-Rel, or Rel-A, which are thereby retained in the cytoplasm (4,5). Following stimulation with TNF␣, and other NF-B agonists, p105 is phosphorylated and then proteolyzed more rapidly by the proteasome (5, 19 -21). This predominantly results in accelerated p105 degradation, rather than increased processing to p50 (21,22). Freed Rel subunits can then translocate into the nucleus to activate gene transcription. A physiological role for p105 in the correct regulation of NF-B has been suggested by the generation of mice lacking its C-terminal (IB-like) half, while still expressing p50 product (23). Such mice display a chronic inflammatory phenotype that correlates with increased nuclear NF-B1 p50 homodimers and involves altered function of T cells, B cells, and macrophages.
It has been shown recently that the mammalian IKK complex immunoprecipitated from transfected cells will phosphorylate recombinant p105 in vitro in a region of the C-terminal PEST domain that appears to be important for TNF␣-induced degradation in HeLa cells (21). Overexpressed IKK2 also promotes the proteolysis of co-transfected p105 in COS cells (24). Together these data are consistent with the hypothesis that the IKK complex directly phosphorylates the PEST domain of p105 to promote its proteolysis after TNF␣ stimulation. However, such in vitro and overexpression experiments are prone to artifact, as exemplified by recent genetic data showing that NIK is not actually required for NF-B activation by proinflammatory cytokines (25) contrary to previous transfection data in cell lines (26). Indeed, the spacing of the putative IKK target serine residues (underlined) in the p105 PEST domain ( 920 DSDSVCDSGVETS) identified in vitro (21) differs considerably from the regulatory phosphorylation sites on IB␣ (see above). Thus it is possible that immunoprecipitated IKK complex does not directly phosphorylate p105 in vitro, but this is mediated by an associated downstream kinase.
In this study, the role of the IKK complex in controlling p105 proteolysis was investigated genetically using knockout cell lines lacking IKK component subunits. These experiments demonstrated that a functional IKK complex is essential for TNF␣ and IL-1␣ to trigger efficiently p105 degradation. Analysis of a panel of point mutants demonstrated that serine 927 in the PEST domain of p105, which is part of a motif related to the IKK phosphorylation site on IB␣, is directly phosphorylated by the IKK complex to regulate its signal-induced degradation. Significantly, this residue does not correspond to any of the three serine residues previously proposed to be important for TNF␣-induced degradation of p105 (21). A simple explanation for this apparent discrepancy is the incorrect designation of this residue as a threonine when the cDNA for p105 was initially cloned and sequenced (27,28). This was confirmed during the preparation of this manuscript by Heissmeyer et al. (29), who also concluded that phosphorylation of serine 927 by the IKK complex regulates signal-induced p105 degradation.

EXPERIMENTAL PROCEDURES
cDNA Constructs and Antibodies-p105 constructs were subcloned into the pcDNA3 expression vector (Invitrogen) for transient mammalian cell expression experiments and into the pMX-1 expression vector (Ingenius) for generation of stably transfected HeLa cell lines. Addition of an N-terminal HA epitope tag to the cDNA encoding human p105 (28) and generation of p105 point mutants were done using the polymerase chain reaction (PCR). PCR was also used to generate a cDNA encoding a GST-p105-(758 -967) fusion protein and its S927A derivative, subcloned into the pGEX-6P vector (Amersham Pharmacia Biotech). For baculovirus expression, IKK1 and IKK2 constructs were tagged at the N terminus with a His 6 tag by PCR and subcloned into the pFastBac vector (Life Technologies, Inc.). DNA templates for PCR of IKK1 (clone 1321982) and IKK2 (clone 3126643) were obtained from Incyte Genomics. All constructs were verified by DNA sequencing. FLAG-IKK2, sub-cloned in the pCMV vector, has been described previously (30).
The following anti-peptide antisera were raised in rabbits to synthetic peptides coupled to keyhole limpet hemocyanin (Pierce): human p105 amino acids 952-967 (anti-h p105-C); murine p105 amino acids 955-971 (anti-m p105-C); human IKK1 amino acids 13-28 (anti-IKK1); and human IKK2 amino acids 572-588 (anti-IKK2). To generate the anti-phospho-Ser 927 antibody, a peptide was synthesized corresponding to residues 922-935 of human p105 in which serine 927 is phosphorylated. After high pressure liquid chromatography purification, this phosphopeptide was coupled to keyhole limpet hemocyanin and injected into rabbits. Anti-IB␣ antibody was kindly provided by Ron Hay (University of St. Andrews, Scotland, UK) and anti-GST mAb by Steve Dillworth (Royal Postgraduate Medical School, London, UK). 12CA5 mAb was used for immunoprecipitation of HA-p105, whereas a high affinity anti-HA mAb (Roche Molecular Biochemicals) was used for its detection in Western blots. Anti-NEMO (IKK␥) and anti-phospho-Ser 32 IB␣ antibodies were purchased from Santa Cruz Biotechnology.
Cell Lines-All cells were cultured in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum, 2 mM L-glutamine, penicillin (100 units/ml), and streptomycin (50 units/ml) and maintained in a rapid growth phase for experiments. Mouse embyronic fibroblasts (MEFs) lacking component subunits of the IKK complex were generously provided by the originating laboratories (see Fig. 1 legend).
To stably transfect HeLa cells (Ohio subline from ECACC) with HA-p105, 7 ϫ 10 5 cells were plated in a 90-mm dish (Life Technologies, Inc.) and, after 18 h in culture, transfected using LipofectAMINE (Life Technologies, Inc.). Transfected cells were cultured for a further 48 h and then selected for neomycin resistance with 1 mg/ml G418 (Life Technologies, Inc.). After 4 weeks, clones were picked manually and then expanded. Expression of HA-p105 was determined by Western blotting.
Pulse-Chase Metabolic Labeling-All pulse-chase metabolic labeling experiments were performed at least twice with similar results. For MEFs, 3 ϫ 10 5 cells were plated per 60-mm dish. After 18 h, cells were washed with phosphate-buffered saline and cultured in methionine/ cysteine-free minimal essential Eagle's medium (Sigma) for 1 h. Cells were pulse-labeled with 2.65 MBq of [ 35 S]methionine/[ 35 S]cysteine (Pro-Mix, Amersham Pharmacia Biotech) for 30 -45 min and chased for the indicated times in complete medium (Dulbecco's modified Eagle's medium plus 2% fetal calf serum) alone (0) or complete medium supplemented with TNF␣ (20 ng/ml; Amersham Pharmacia Biotech) or IL-1␣ (4 ng/ml; R & D Systems). Cells were lysed in buffer A (31) and endogenous p105 immunoprecipitated as described previously (22) using a 1:1 mixture of anti-h p105-C and anti-m p105-C (10 l of antiserum total per immunoprecipitate).
NIH-3T3 cells were transiently transfected using LipofectAMINE (Life Technologies, Inc.). Pulse-chase metabolic labeling and immunoprecipitation were carried out as described previously (22). Labeled bands were quantified by laser densitometry (Calibrated Imaging Densitometer, Bio-Rad). To analyze HA-p105 proteolysis in stably transfected HeLa clones, cells were plated at 6 ϫ 10 5 per well of a 6-well plate (Life Technologies, Inc.). After 18 h of culture, pulse-chase metabolic labeling was carried out as for MEFs. HA-p105 was isolated by immunoprecipitation with 12CA5 anti-HA mAb after lysis with buffer A supplemented with 0.1% SDS and 0.5% deoxycholate.
Analysis of p105 Phosphorylation-To analyze in vivo phosphorylation of p105 on serine 927, 5 ϫ 10 6 HeLa cells were plated per 100-mm dish. After 18 h in culture, cells were pretreated with 20 M MG132 proteasome inhibitor for 30 min (Biomol Research Labs) and then stimulated for the indicated times with IL-1␣ (4 ng/ml), TNF␣ (20 ng/ml), or control medium. Endogenous p105 was then immunoprecipitated using anti-h p105-C antiserum, following extraction in buffer A, and Western-blotted with anti-phospho-Ser 927 antiserum. Bovine serum albumin (Sigma) was used as a blocking agent for blots.
For in vitro phosphorylation experiments with endogenous IKK complex, control or TNF␣-stimulated HeLa cells (7 ϫ 10 5 ) were lysed in Buffer A and immunoprecipitated using anti-NEMO antibody. Immunoprecipitates were washed four times in Buffer A and once in kinase buffer (25 mM Tris, pH 7.5, 5 mM ␤-glycerophosphate, 2 mM dithiothreitol, 0.1 mM sodium vanadate, 10 mM MgCl 2 ). Immunoprecipitates were then incubated at room temperature in 50 l of kinase buffer plus 40 M ATP and 5 g of purified GST-p105-(758 -967) or GST-p105-(758 -967)-(S928A) fusion protein. The reaction was stopped by addition of 50 l of 2ϫ Laemmli sample buffer and proteins Western blotted with antiphospho-Ser 927 antiserum.
His 6 -IKK1 and His 6 -IKK2 were expressed in Sf9 insect cells by baculovirus infection. Proteins were isolated from 1 liter of cell pellet extracted with 50 ml of lysis buffer (20 mM Tris, pH 8.0, 137 mM NaCl, 10% glycerol, and 1.0% Triton X-100) plus protease inhibitor mix (Roche Molecular Biochemicals). Centrifuged lysate was applied to a 5-ml chelating Sepharose HiTrap column (Amersham Pharmacia Biotech) that was pretreated with NiCl 2 and equilibrated in 50 mM HEPES, pH 7.5, 300 mM NaCl. The column was then washed with 50 mM imidazole and bound protein eluted with 250 mM imidazole. The protein was buffer exchanged into 50 mM HEPES, pH 7.5, 10% glycerol, 5 mM dithiothreitol using a 5-ml Desalt HiTrap column (Amersham Pharmacia Biotech) and stored at Ϫ80°C. Protein was Ͼ90% pure as judged by Coomassie Blue staining of SDS-PAGE gels (data not shown). For in vitro kinase assays with baculovirus IKK-1/2, 25 ng of recombinant IKK protein was used per reaction which were carried out as described above.

Analysis of p105 Degradation in Fibroblast Cells Lacking
Component Subunits of the IKK Complex-A genetic approach was taken to investigate the role of the IKK complex in p105 degradation induced by physiological stimulation to avoid potential problems associated with the use of dominant negative kinase mutants (see the Introduction). To do this, a series of pulse-chase metabolic labeling experiments were carried out in embryonic fibroblast (MEF) cells isolated from the indicated knockout mice to determine p105 turnover after cytokine stimulation.
In cells lacking either IKK1 (32) or IKK2 (33), both TNF␣ and IL-1␣ increased the proteolysis of p105 to a similar degree to their wild type counterparts (Fig. 1, A and B). Control experiments also demonstrated that neither IKK1 nor IKK2 was absolutely essential for TNF␣ and IL-1␣ to trigger degradation of IB␣, although the extent of IB␣ degradation was markedly decreased in cells lacking IKK2 (Fig. 1, A and B). In contrast, cytokine-induced degradation of p105 was significantly reduced, and IB␣ was completely blocked, in MEF cells lacking NEMO expression ((34) Fig. 1C). Control experiments confirmed that both TNF␣ and IL-1␣ stimulation of NEMO-negative cells promoted phosphorylation of p38 mitogen-activated protein kinase, indicating that the TNF␣ and IL-1 signaling pathways were still intact in these cells (data not shown). Since NEMO is essential for IKK activation by pro-inflammatory cytokines (13,34,35), these data demonstrated that a functional IKK complex is required for both TNF␣ and IL-1␣ to trigger efficiently p105 proteolysis. IKK1 and IKK2, however, are redundant in this signaling pathway.
The PEST Domain of p105 Contains a Sequence Highly Related to the IKK Target Sequence of IB␣-The experiments with IKK knockout cell lines suggested that p105 might be a direct target of IKK. Interestingly, a data base search revealed a conserved motif in p105 that is similar to the IKK complex phosphorylation site in IB␣. Fig. 2A compares an alignment of residues 918 -935 from the PEST domain of human p105 with p105 sequences from other species. This region contains two overlapping sequences related to the IKK target sequence in I␤␣ that are identical in the p105 sequences from human, mouse, and rat. In chicken p105, however, only the second of these overlapping motifs are identical, and the first differed at two positions from the human p105 sequence ( Fig. 2A). Interestingly, the regulatory IKK phosphorylation site at the N terminus of IB␣ is identical between human, mouse, rat, and chicken sequences (36).
In addition, a discrepancy was noted at residue 927, which is a serine in three of the published human p105 sequences (Gen-Bank TM accession numbers: 227315, AAF35232, and XP003398), and also in the corresponding residue in mouse, rat, and chicken p105. However, in three other published human p105 sequences there is a threonine at residue 927 (Gen-Bank TM accession numbers: P19838, AAA36361, and AAA94041). To confirm directly the sequence in this region of human p105, clones encoding p105 cDNA were obtained from two different laboratories and sequenced over the DNA region encoding amino acids 918 -935. Both sequences predicted a serine at amino acid 927, although one of these had been reported previously to encode a threonine at this position (28). This difference is accounted for by a single base pair error in the original published DNA sequence. DNA was also extracted from primary human lymphocytes from four donors and sequenced. Again these sequences all predicted a serine at amino acid 927. Together these analyses clearly indicated that residue 927 in human p105 is a serine rather than a threonine.
Human, mouse, rat, and chicken p105, therefore, all contain a sequence (D n SGVET nϩ5 ) that is closely related to the IB␣ regulatory sequence (D n SGLDS nϩ5 ) and contains conservative substitutions at residues n ϩ 3 and n ϩ 4 but in which the second serine at n ϩ 5 is replaced with a threonine (Fig. 2, A and B). Anti-p105C immunoprecipitates (anti-p105C Ip) were resolved by 10% SDS-PAGE and revealed by fluorography (upper panels). In parallel experiments, knockout and control fibroblasts were stimulated for the indicated times with TNF␣, IL-1␣, or control medium, and lysates were Western-blotted for IB␣ and the relevant IKK subunits (lower panels).
Identification of Serine Residues Involved in IKK2-mediated p105 Proteolysis-A series of serine or threonine to alanine mutants were generated in the region of the human p105 PEST domain containing the two potential overlapping IKK target sites (Fig. 2C). Overexpressed IKK2 triggers proteolysis of cotransfected p105 (24). Accordingly, to investigate whether any of the p105 mutations affected IKK2-promoted p105 proteolysis, 3T3 fibroblasts were co-transfected with plasmids encoding HA epitope-tagged p105 together with IKK2 or empty vector. HA-p105 turnover was then assessed by immunoprecipitating HA-p105 from lysates of pulse-chase metabolically labeled cells and resolving isolated protein by SDS-PAGE. Overexpressed IKK2 dramatically decreased the half-life of pulse-labeled HA-p105 compared with empty vector control (Fig. 3, A and B). The S921A, S923A, and T931A mutants of HA-p105 were all degraded similar to wild type HA-p105 in response to IKK2 co-expression (Fig. 3B). HA-p105(S932A) was also degraded when co-expressed with IKK2 but consistently with delayed kinetics compared with wild type HA-p105 (Fig. 3B). However, IKK2 had little detectable effect on the proteolysis of HA-p105(S927A) (Fig. 3, A and B). Thus serine 927 plays a crucial role in controlling the proteolysis of p105 triggered by IKK2 overexpression.
TNF␣ and IL-1 Cannot Trigger Proteolysis of an S927A Mutant of p105-It was important to demonstrate that serine 927 is important for proteolysis of p105 induced by physiological stimulation with cytokines, in addition to its essential role in proteolysis triggered by IKK2 overexpression. To this end, HeLa cells were transfected with plasmids encoding HA-p105 or HA-p105(S927A) and stable clones isolated after G418 selection. A pulse-chase experiment was carried out with a representative clone expressing levels of HA-p105(S927A) similar to that of a control clone expressing HA-p105 (data not shown). Both TNF␣ and IL-1␣ increased the rate of turnover of HA-p105 (Fig. 4A). In contrast, neither cytokine affected the turnover of HA-p105(S927A), although both induced a p105 mobility shift in SDS-PAGE which was probably due to phosphorylation on amino acids other than serine 927 (20,22). The turnover of stably transfected HA-p105 in unstimulated cells was consistently found to be much greater than endogenous p105 in parental HeLa cells (data not shown), as found previously by Heissmeyer et al. (21). However, this basal turnover was also blocked in the HA-p105(S927A) mutant (Fig. 4A). Control experiments demonstrated that IB␣ was degraded following stimulation with either TNF␣ or IL-1␣ in the HA-p105(S927A) clone, similar to the HA-p105 wild type clone (Fig.  4B). The TNF␣ and IL-1␣ signaling pathways controlling the IKK complex were, therefore, intact in the HA-p105(S927A) clone. Similar results were obtained with an independently derived HA-p105(S927A) clone (data not shown). Thus introduction of an S927A mutation into p105 completely blocks its proteolysis induced by two different pro-inflammatory cytokines, confirming the importance of serine 927 in regulating signal-induced p105 degradation.
TNF␣ and IL-1 Induce Rapid Phosphorylation of p105 on Serine 927-The previous experiments indicated an important role for serine 927 of p105 in its signal-induced proteolysis. However, it was important to obtain direct evidence that serine 927 was phosphorylated in vivo, as it was possible that the S927A mutation might mediate its inhibitory effects on p105 proteolysis via a structural alteration, rather than by preventing serine 927 phosphorylation. To investigate this possibility, an antibody was raised against a synthetic peptide corresponding to residues 921-933 of human p105 in which serine 927 was phosphorylated (Fig. 2, A and B).
To test this reagent, 3T3 cells were transfected with plasmids encoding wild type HA-p105 or point mutants thereof plus and minus an IKK2 plasmid and cultured for 24 h. Cells were incubated with MG132 inhibitor for the last 4 h of culture to block proteasome-mediated proteolysis (6). HA-p105 immunoprecipitated from cells co-expressing IKK2 was clearly recognized in Western blots as two distinct bands by the anti-phospho-Ser 927 antibody (Fig. 5A, upper panel). The slower mobility band was probably caused by IKK2-induced phosphorylation of HA-p105 on multiple amino acids. Similar results were obtained with the S921A, S923A, T931A, and S932A mutants. However, neither of the bands were detected in cells co-transfected with HA-p105(S927A), although HA-p105(S927A) protein was clearly present in the immunoprecipitates as revealed by immunoblotting with anti-h p105-C antibody (Fig. 5A, lower  panel). In addition, the IKK2-induced bands detected in cells co-transfected with HA-p105 were completely lost when blots were incubated with anti-phospho-Ser 927 antibody plus Ser 927 phosphopeptide but not with the unphosphorylated peptide (data not shown). Together these data confirmed the specificity of the anti-phospho-Ser 927 antibody and demonstrated that IKK2 overexpression induces phosphorylation of co-transfected HA-p105 on serine 927.
To investigate whether pro-inflammatory cytokines induce phosphorylation of endogenous p105 on serine 927, HeLa cells were preincubated with MG132 for 30 min to block the proteasome and then stimulated for the indicated times with either TNF-␣ or IL-1␣ (Fig. 5B). p105 was isolated from cell lysates by immunoprecipitation and then immunoblotted with the antiphospho-Ser 927 antibody. TNF␣ stimulation induced rapid ser- ine 927 phosphorylation that reached a maximum at 15 min and then gradually declined. Stimulation with IL-1␣ also induced phosphorylation of p105 on Ser 927 but with delayed kinetics (peaking at 30 -60 min) and reduced amplitude compared with TNF␣. Immunoblotting with anti-h p105C antibody confirmed that equal amounts of p105 were present in all of the immunoprecipitates. Lysates were re-immunoprecipitated with an anti-IB␣ antibody, and isolated protein was immunoblotted with an antibody specific for phospho-Ser 32 of IB␣. The kinetics of IB␣ Ser 32 phosphorylation (Fig. 5B, lower panels) were very similar to those of p105 phosphorylation on Ser 927 , and the TNF␣ signal was greater than that detected with IL-1␣, as also found with p105 Ser 927 phosphorylation.
The data in this section confirm that p105 serine 927 is phosphorylated under stimulatory conditions that promote p105 proteolysis and strongly suggest that the inhibitory effects of the S927A mutation on signal-induced p105 proteolysis are due to removal of a regulatory phosphorylation site.
The IKK Complex Phosphorylates a Recombinant p105 Fusion Protein on Serine 927 in Vitro-Previous research has established that the IKK-1/2 complex preferentially phosphorylates serine residues over threonine. A mutant IB␣ in which the regulatory phosphorylation sites (Ser 32 /Ser 36 ) are mutated to threonine is phosphorylated and degraded at significantly reduced levels in stimulated cells compared with the wild type protein (11,30,37). The experiments in the previous sections indicated that IKK2 overexpression triggered degradation of co-transfected p105 as a result of inducing phosphorylation of p105 on serine 927 (Figs. 3 and 5A). To initially investigate whether the IKK complex might directly phosphorylate p105 on serine 927 to regulate its degradation, a mutant was generated in which residue 927 was altered to threonine. Similar to HA-p105(S927A), HA-p105(S927T) was not degraded when coexpressed with IKK2 (Fig. 6A). In contrast, IKK2 co-expression induced the rapid degradation of HA-p105 wild type, as expected. These data indicate that the kinase that phosphorylates p105 at residue 927 is serine-specific, consistent with the hypothesis that it might correspond to the IKK complex.
To investigate directly whether the IKK complex could phosphorylate p105 serine 927, a GST fusion protein was purified from Escherichia coli transformed with a plasmid encoding the C terminus of p105 (residues 758 -967; GST-p105-(758 -967)) and a corresponding mutant in which the residue equivalent to p105 serine 927 was mutated to alanine (GST-p105-(758 -967)(S927A)). HeLa cells were stimulated with TNF␣ for 15 min, or left unstimulated, and the endogenous IKK complex was isolated by immunoprecipitation using an anti-NEMO antibody. An in vitro kinase assay was then carried out using purified GST-p105-(758 -967) protein as a substrate, and serine 927 phosphorylation was assessed by Western blotting using the anti-phospho-Ser 927 antibody. Immunopurified IKK complex phosphorylated the GST fusion protein on serine 927, and this was dramatically increased after TNF␣ stimulation (Fig. 6B), which activates the IKK complex (15). As expected, no signal was detected when GST-p105-(758 -967)(S927A) was used as a substrate for anti-NEMO immunoprecipitates from TNF␣-stimulated cells (lane 5). In addition, control immunoprecipitates with non-immune serum (lanes 1 and 2) did not isolate any p105 Ser 927 kinase activity either from control or TNF␣-stimulated cells. These data indicate that the endogenous IKK complex isolated from HeLa cells directly phosphorylates serine 927 of p105.
IKK-1 and IKK-2 were also individually produced using a baculovirus expression system and isolated by affinity purification to over 90% purity (data not shown). These proteins each displayed constitutive kinase activity in the absence of NEMO, as noted previously (12). In vitro kinase assays demonstrated that both His 6 -IKK-1 and His 6 -IKK-2 phosphorylated the GST-p105-(758 -967) fusion protein on Ser 927 (Fig. 6C). No signal was detected with the control GST-p105-(758 -967)(S927A) fusion protein confirming the specificity of the anti-phospho-Ser 927 antibody. Thus, both IKK-1 and IKK-2 can directly phosphorylate serine 927 of p105, consistent with ability of TNF␣ and IL-1␣ to induce p105 degradation in IKK1 Ϫ/Ϫ and IKK2 Ϫ/Ϫ MEFs (see Fig. 1). The data in this section, therefore, show that p105 serine 927 is a direct target of the IKK complex.

DISCUSSION
The combined genetic and biochemical data in this study demonstrate that the mammalian IKK complex directly regulates the signal-induced degradation of NF-B1 p105 (12). These data strengthen the notion that the IKK complex plays a central role in the regulation of NF-B activation by inducing degradation of the major IB proteins (IB␣, IB␤, and p105) in mammals (15,38).
Activation of the IKK complex by TNF␣ and IL-1 involves distinct upstream signaling intermediates (15, 38 -41). However, both of these cytokines induce the IKK complex to phosphorylate serine 927 of NF-B1 p105 to trigger its degradation. Similar to IB␣ (1, 15), therefore, the signal-induced degradation of p105 by different stimuli involves a common phosphorylation-based mechanism, and the IKK complex integrates the signals from these two different signaling pathways to trigger p105 degradation. The signal-induced proteolysis of p105, therefore, is controlled at the level of IKK complex activation and perhaps also recruitment of the IKK complex to p105 (29). Interestingly, basal turnover of p105 is also blocked by mutation of serine 927 (Fig. 4A) and, therefore, may also be regulated by the IKK complex.
A previous study from this laboratory demonstrated that the C terminus of NF-B1 p105 is stably associated at high stoichiometry with the mitogen-activated protein 3-kinase, TPL-2/ Cot (22). Overexpression of TPL-2/Cot induces phosphorylation and degradation of co-expressed p105 in 3T3 cells. TPL-2/Cot, however, cannot directly phosphorylate p105 on serine 927 in vitro but does induce serine 927 phosphorylation when coexpressed with HA-p105 in 3T3 cells. 2 Since TPL-2/Cot over- expression has been shown to activate the IKK complex (42), it is likely that TPL-2/Cot regulates serine 927 phosphorylation and subsequent degradation of p105 via the IKK complex. This possibility is currently being investigated.
An earlier study showed that the IKK complex could phosphorylate p105 in vitro (21). Analysis of internal deletion mutants revealed that the major in vitro phosphorylation sites resided between residues 920 and 936 which contained three serine residues according to original sequences of p105 published by this and other laboratories. A triple mutant in which these serine residues were mutated to alanine (S921A,S923A,S932A) was no longer phosphorylated by IKK in vitro and was not degraded in response to TNF␣ stimulation when expressed in HeLa cells as a C-terminal p105 fragment. Surprisingly, in the present study, it was found that individual mutation of each of these residues to alanine did not block IKK2-triggered proteolysis of HA-p105 in co-transfected 3T3 cells, whereas an S927A p105 mutation completely prevented HA-p105 proteolysis promoted by IKK2 co-expression. A serine-to-thre-onine mutation at residue 927 was also found to block IKK2mediated proteolysis of HA-p105 (Fig. 6A), consistent with known preference of the IKK complex for serine residues over threonine (11,30,37,43). The most likely explanation for the discrepancy between the results of the present study and those of Heissmeyer et al. (21) is that a 927 serine to threonine mutation was inadvertently introduced into the triple mutant of p105 during PCR, which was presumably performed on the assumption that residue 927 was a threonine. This hypothesis was confirmed in a more recent paper (29) from the same laboratory published while this article was in preparation.
Heissmeyer et al. (29) demonstrate a critical role for serine 927 in IKK2-induced p105 proteolysis and a minor role for serine 923 in transiently transfected 293 cells. By analysis of serine/threonine to alanine mutants of p105, it is also shown that purified baculovirus IKK2 phosphorylated a p105 fusion protein on both of these residues in vitro (29). The effects of mutating serine 927 to alanine on p105 proteolysis are consistent with the present study, which also extends these observations by directly demonstrating in vivo phosphorylation of this residue (Fig. 5) and its critical importance in cytokine-induced p105 degradation (Fig. 3A). A regulatory role for serine 923 in p105 proteolysis was not observed in the present study, as HA-p105(S923A) is degraded similarly to wild type HA-p105 when co-expressed with IKK2 in 3T3 cells (Fig. 3B). This discrepancy may relate to use of different cell lines for transient transfections. It remains to be determined, however, whether p105 serine 923 is actually phosphorylated in vivo when coexpressed with IKK2 or after cytokine stimulation.
Mutation of serine 932 to alanine consistently delayed, but did not completely block, HA-p105 proteolysis triggered by IKK2 overexpression in 3T3 cells (Fig. 3B), suggesting that this residue may also play a regulatory role in p105 degradation. Serine 932 is not phosphorylated by IKK2 in vitro (29), and its mutation to alanine does not affect IKK2-mediated phosphorylation of serine 927 in vivo (Fig. 5A). Thus it is possible that a serine 932 kinase, distinct from the IKK complex, is also involved in the regulation of p105 proteolysis. A regulatory role for this residue may be cell type-specific, however, as Heissmeyer et al. (29) did not detect any inhibitory effect of mutation of this residue on IKK2-induced p105 degradation in 293 cells.
The Asp-Ser 927 -Gly-Val-Glu-Thr 931 motif in the PEST domain of p105 is closely related to the N-terminal IKK target sequence of IB␣ (Asp-Ser 32 -Gly-Leu-Asp-Ser 36 ) (8,9,11,30,37), except for the substitution of a serine for threonine at the residue 931 of p105, which corresponds to serine 36 of IB␣. Previous experiments demonstrated that mutation of serine 36 to threonine only slightly reduces IB␣ degradation in response to TNF␣ stimulation (11). However, mutation of serine 32 of IB␣ to threonine has a more dramatic inhibitory effect. In vitro experiments have also established that the IKK complex can phosphorylate serine 32 in an I␤␣ fusion protein containing an Ser 36 to Thr mutation (12). Together, these data are consistent with the findings in the present study showing a crucial role for serine 927 of the Asp-Ser 927 -Gly-Val-Glu-Thr motif in the regulation of p105 proteolysis by IKK2 overexpression and its direct phosphorylation by the IKK complex in vitro. Western blot analyses indicate that p105 is only partially degraded after TNF␣ stimulation of HeLa cells, whereas degradation of IB␣ is complete (data not shown). The presence of a threonine, rather than a serine, at residue 931 of p105 may contribute to the inefficient proteolysis of p105 triggered by cytokine stimulation.
In preliminary experiments, it has been shown that serine 927 is essential for signal-induced ubiquitination of p105 (data not shown). It is likely that this reflects a requirement for serine 927 phosphorylation to recruit an ubiquitin-protein isopeptide ligase to p105, thereby facilitating its subsequent then chased for the times indicated. Anti-HA immunoprecipitates were resolved by 7.5% SDS-PAGE and revealed by fluorography. p105 was quantified by laser densitometry. B, HeLa cells were stimulated for 15 min with TNF␣ or left unstimulated. The endogenous IKK complex was isolated from cell lysates by immunoprecipitation with an anti-NEMO antibody. Control immunoprecipitates were carried out with non-immune rabbit serum (NRS). Immunoprecipitates were tested for their ability to phosphorylate in vitro GST-p105-(758 -967) or GST-p105-(758 -967)(S927A) fusion proteins (as indicated). Phosphorylation was determined by Western blotting kinase reaction mixtures with antiphospho-Ser 927 antibody (upper panel). Blots were then stripped and reprobed sequentially with anti-NEMO antibody (middle panel) and anti-GST mAb (lower panel). C, in vitro kinase assays were carried out with baculovirus-produced purified His 6 -IKK1 or His 6 -IKK2 (25 ng), using as substrates either GST-p105-(758 -967) or GST-p105-(758 -967)(S927A) fusion protein. Phosphorylation was assessed by Western blotting kinase reactions with anti-phospho-Ser 927 antibody (upper panel). Equal loading of the fusion proteins was confirmed by reprobing blots with anti-GST mAb (lower panel). ubiquitination and degradation by the proteasome. Recent published data have indicated that ␤TrCP, a component subunit of an SCF-type ubiquitin-protein isopeptide ligase, mediates the ubiquitination of phosphorylated p105 (24), similar to its established role in IB␣ degradation (44 -46). Deletion of p105 residues 917-933 prevents both the interaction of ␤TrCP with phospho-p105 and the ubiquitination of phospho-p105 in vitro (24). Furthermore, a quadruple mutant of p105, encoding serine to alanine mutations at residues 921, 923, 927, and 933, is no longer ubiquitinated when co-transfected with IKK2 and ␤TrCP (29). It will be important in future studies to determine whether serine 927 phosphorylation is required for ␤TrCPmediated p105 ubiquitination and in particular whether phospho-Ser 927 is the binding site for ␤TrCP on p105.
The Drosophila genome contains three NF-B/Rel family members: Dorsal, Dorsal-related immunity factor, and Relish (47). Dorsal-related immunity factor and Dorsal are retained in the cytoplasm of unchallenged flies by binding to the IB␣ homolog Cactus (48). Cactus is degraded in response to an immune stimulus via a Toll signaling pathway, releasing associated Dorsal-related immunity factor and Dorsal to activate the expression of anti-fungal peptide genes (49 -51). Relish contains an N-terminal RHD and an IB-like ankyrin-repeat in its C terminus similar to mammalian p105 (52). During antibacterial immune responses, Relish is cleaved by DREDD caspase to release an N-terminal fragment that translocates into the nucleus to promote expression of anti-bacterial peptides (53). Drosophila IKK2 (DmIKK␤) and NEMO (DmIKK␥/ Kenny) are essential for this signaling pathway but not for signal-mediated degradation of Cactus during anti-fungal responses (54 -56). Since Cactus degradation is triggered by phosphorylation at its N-terminal regulatory domain followed by ubiquitination similar to IB␣ (46), a separate IKK complex may be involved in regulating Cactus proteolysis (54).
Thus it appears that distinct kinases regulate the degradation of different IBs in Drosophila. This contrasts with the situation in mammals where the IKK complex is now known to regulate directly the degradation of three IB proteins: IB␣, IB␤ (12), and NF-B1 p105 (this paper). IB⑀ may also be directly phosphorylated by the IKK complex (57,58). However, mammalian NF-B2 p100 is not phosphorylated by the IKK complex in vitro (21), and recent data have indicated that signal-induced proteolysis of this protein is regulated by a distinct kinase (60).