The IκB Kinase (IKK) Complex Is Tripartite and Contains IKKγ but Not IKAP as a Regular Component*

A critical step in the activation of NF-κB is the phosphorylation of IκBs by the IκB kinase (IKK) complex. IKKα and IKKβ are the two catalytic subunits of the IKK complex and two additional molecules, IKKγ/NEMO and IKAP, have been described as further integral members. We have analyzed the function of both proteins for IKK complex composition and NF-κB signaling. IKAP and IKKγ belong to distinct cellular complexes. Quantitative association of IKKγ was observed with IKKα and IKKβ. In contrast IKAP was complexed with several distinct polypeptides. Overexpression of either IKKγ or IKAP blocked tumor necrosis factor α induction of an NF-κB-dependent reporter construct, but IKAP in addition affected several NF-κB-independent promoters. Whereas specific down-regulation of IKKγ protein levels by antisense oligonucleotides significantly reduced cytokine-mediated activation of the IKK complex and subsequent NF-κB activation, a similar reduction of IKAP protein levels had no effect on NF-κB signaling. Using solely IKKα, IKKβ, and IKKγ, we could reconstitute a complex whose apparent molecular weight is comparable to that of the endogenous IKK complex. We conclude that while IKKγ is a stoichiometric component of the IKK complex, obligatory for NF-κB signaling, IKAP is not associated with IKKs and plays no specific role in cytokine-induced NF-κB activation.

From the ‡Max-Delbrü ck-Centrum for Molecular Medicine, Robert-Rössle-Str. 10,13125 Berlin, Germany, the ¶Department of Biology, Boehringer Ingelheim Pharmaceuticals, Ridgefield, Connecticut 06877-0368, ʈAtugen AG, Robert-Rössle-Str. 10,13125 Berlin, Germany, and **Micromet GmbH, Am Klopferspitz 19, 82152 Martinsried, Germany A critical step in the activation of NF-B is the phosphorylation of IBs by the IB kinase (IKK) complex. IKK␣ and IKK␤ are the two catalytic subunits of the IKK complex and two additional molecules, IKK␥/NEMO and IKAP, have been described as further integral members. We have analyzed the function of both proteins for IKK complex composition and NF-B signaling. IKAP and IKK␥ belong to distinct cellular complexes. Quantitative association of IKK␥ was observed with IKK␣ and IKK␤. In contrast IKAP was complexed with several distinct polypeptides. Overexpression of either IKK␥ or IKAP blocked tumor necrosis factor ␣ induction of an NF-Bdependent reporter construct, but IKAP in addition affected several NF-B-independent promoters. Whereas specific down-regulation of IKK␥ protein levels by antisense oligonucleotides significantly reduced cytokinemediated activation of the IKK complex and subsequent NF-B activation, a similar reduction of IKAP protein levels had no effect on NF-B signaling. Using solely IKK␣, IKK␤, and IKK␥, we could reconstitute a complex whose apparent molecular weight is comparable to that of the endogenous IKK complex. We conclude that while IKK␥ is a stoichiometric component of the IKK complex, obligatory for NF-B signaling, IKAP is not associated with IKKs and plays no specific role in cytokine-induced NF-B activation.
NF-B transcription factors play a pivotal role in many cellular processes such as inflammation, immune response, cell proliferation, and apoptosis (1)(2)(3)(4)(5). The prototype of the NF-B family is a heterodimer of the p50 and p65 (RelA) subunits. IB proteins (IB␣, IB␤, IB⑀, p105, and p100) retain NF-B in an inactive form in the cytoplasm. A conserved ankyrin repeat domain in these inhibitors masks nuclear translocation signals contained in the Rel homology domain of NF-B.
In response to multiple stimuli, including TNF␣, 1 IL-1␤, phorbol ester, and lipopolysaccharides, NF-B is liberated from IB molecules and translocates to the nucleus (6). This critical step of NF-B activation is initiated by phosphorylation of IB proteins at conserved amino-terminal serine residues, e.g. at serines 32 and 36 of IB␣ or serines 19 and 23 of IB␤. Phosphorylated IBs are bound by a ␤TrCP containing ubiquitin ligase (E3) complex, polyubiquitinated and subsequently degraded by the 26 S proteasome (7).
Most NF-B-inducing stimuli trigger activation of an IB kinase (IKK) complex with a high apparent molecular mass of 700 -900 kDa (8,9), which has specificity for the amino-terminal phosphoacceptor sites in IB␣ or -␤. The kinase complex contains two catalytic subunits termed IKK␣ (IKK1) and IKK␤ (IKK2) (8, 10 -13). IKK␣ and IKK␤ are related molecules of 85 and 87 kDa, respectively, with an overall identity of about 44%. Both contain an NH 2 -terminal kinase domain, a leucine zipper and a COOH-terminal helix-loop-helix motif. IKK␣ and IKK␤ form homo-or heterodimers via their leucine zipper (for review, see Ref. 14). Both kinases are stimulated by proinflammatory cytokines and their activation kinetics match that of IB␣ phosphorylation. Highly purified recombinant IKK␣ and IKK␤ can phosphorylate IB␣ and IB␤ directly at the correct sites, thus no further downstream kinases are required for IB phosphorylation (15,16). IKK␣ and IKK␤ also inducibly phosphorylate the NF-B precursor protein p105 at three carboxylterminal serines and thereby trigger proteolysis of the precursor (17). The IKK complex appears to contain a IKK␣/␤ heterodimer (9), although in some cell types IKK␤ homodimers are found as well (18). The recent generation of IKK␣-or IKK␤-deficient mice has established the requirement of IKK␤ for activation of NF-B by proinflammatory stimuli (19 -21). In contrast, IKK␣ was found to be dispensable for these stimuli but was essential for morphogenic functions, including differentiation and proliferation of epidermal keratinocytes and skeletal development (22)(23)(24). The IKK␣ knock-out model also demonstrated that the signal responsiveness and activity of the resulting IKK␤ homodimer in these animals is fully functional. A predominant role of IKK␤ for proinflammatory signaling is also evident from the observation that mutation of two amino acids in the activation loop of IKK␤, but not in IKK␣, blocks IKK activation by NIK or cytokines (25).
In addition to the two kinases, the IKK complex has been reported to contain regulatory subunits. IKK␥ (NEMO, IKKAP1) has been obtained by complementation cloning (26) and by microsequencing of the purified protein (18,27). Murine IKK␥ (NEMO) restored the defect of mutant cell lines which had lost the ability to activate NF-B (26). IKK␥ has an extended coiled-coil structure prediction, forms dimers and trimers in vitro (26,27), and directly binds to IKK␤ but not to IKK␣ (27). IKK␥ is required for activation of NF-B by TNF␣, IL-1␤, lipopolysaccharide, phorbol 12-myristate 13-acetate, double stranded RNA, or the human T-cell lymphotrophic virus (HTLV-1) Tax, as shown with IKK␥-deficient cells or by overexpression of antisense cDNA (26,27). Most, if not all of the cellular IKK␣ and -␤ seems to be bound to IKK␥, the absence of which results in a shift of the apparent molecular mass of the IKK complex from 800 -900 to 300 -400 kDa (26). A carboxylterminal truncated IKK␥, which still binds to IKK␣ and -␤, acts as an inhibitor of cytokine-induced, but not of basal IKK kinase activity, when overexpressed (27). The COOH-terminal domains containing a leucine zipper and potential zinc finger could thus be interaction sites for signal transmitting molecules which activate IKK␣ or -␤ (18,27). IKK␥ was also isolated in a yeast two-hybrid screen using the adenovirus-encoded E3-14.7K protein as a bait and was shown to interact with RIP and NIK in transfected cells (28). The same adenovirus protein also interacts with Fip2, which shares 40% similarity with IKK␥ and also contains coiled coils and a leucine zipper (29).
A further IKK-associated protein (IKAP) was isolated by affinity purification of the IKK complex using immobilized IB␣ (30). IKAP is a 150-kDa protein with an amino-terminal WD40-like repeat domain. IKAP was found to co-elute with large molecular size range proteins (900 kDa) and co-eluted with IKK activity. IKAP co-purified with IKK␣, IKK␤, NIK, RelA, IB␣, and further proteins of 105, 100, 82, 80, 65, and 58 kDa. IKAP was reported to directly and independently interact with recombinant or transfected IKK␣, IKK␤, and NIK. The sequestration of IKKs and NIK suggested a function of IKAP as a scaffold protein (30).
It is not clear whether all endogenous IKK complexes associate with IKAP or only a subset of IKKs. Furthermore, it is not known whether IKK␥ and IKAP are present in the same IKK complexes and how many other components are commonly associated with heterodimers of IKK␣ and IKK␤. In this study we have analyzed the composition of cellular IKK␣⅐IKK␤ complexes. We come to the conclusion that with regard to composition the major IKK complex is only tripartite and consists exclusively of IKK␣, -␤, and -␥. With in vitro reconstitution experiments we show that an IKK␣⅐␤⅐␥ complex displays a gel filtration profile which, like that of endogenous IKK complexes, corresponds to a large size of more than 800 kDa relative to standard proteins. Various protein-protein interaction and functional assays demonstrate that cellular IKK complexes do not contain IKAP as an intrinsic component. IKAP appears as part of a novel complex containing additional proteins of 100, 70, 45, and 39 kDa. Overexpression of IKAP interferes with the activity of a set of different NF-B-dependent as well as independent reporter genes, suggesting a function in a more general gene expression mechanism.
Transfection and Luciferase Assay-293 cells were transiently transfected by the calcium phosphate precipitation method as described previously (17). DNA constructs and amounts are indicated in the figure legend. Stimulation using 50 ng/ml TNF␣ was carried out 6 h prior to lysis. Cells were lysed 24 h after transfection and luciferase measurement was done with the dual luciferase reporter kit (Promega) according to the manufacturer's protocol. For transfection of antisense oligonucleotides (GeneBlocs) HeLa cells were plated in 96-well dishes at 4000 cells/well the evening before transfection. Cells were transfected in triplicate with antisense or control oligonucleotides for up to 72 h in the presence of serum using a lipid-based delivery system (Atugen AG). Transfection efficiency was greater than 80% (data no shown). Cells were harvested and RNA or protein extracts were prepared 48 h after transfection. Relative amounts of mRNA were determined by Real Time TaqMan TM PCR analysis using the ABI Prism 7700 system (PE Applied Biosystems).
Extracts, Electrophoretic Mobility Shift Assay, and Western Blotting-Whole cell extracts were analyzed by electrophoretic mobility shift assay and Western blotting essentially as described previously (31). For the preparation of cytoplasmic, nuclear and chromatin extracts HeLa cells were washed with phosphate-buffered saline and swollen in buffer A (1 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 1 mM dithiothreitol (DTT) plus protease inhibitors, 0.4 mM Pefabloc, 1 g/ml aprotinin, 1 g/ml leupeptin, and 1 g/ml pepstatin A) for 10 min on ice. 0.15% Nonidet P-40 was added, thoroughly mixed, and spun down in a microcentrifuge. The supernatant was centrifuged for 10 min at 14,000 rpm and after addition of 10% glycerol used as cytoplasmic extract. The nuclei were shaken for 15 min in buffer C (20 mM HEPES, pH 7.9, 20% glycerol, 0.42 M NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 1 mM DTT, plus protease inhibitors (see above)) and after centrifugation for 10 min at 14,000 rpm the supernatant was used as a nuclear extract. The pellet was washed once with ϳ10 volumes of buffer C and directly taken up and boiled for 15 min in SDS loading buffer to elute proteins that are tightly bound to chromatin.
All gel filtration chromatography was carried out on a Superose 6 column (Amersham Pharmacia Biotech). 500-l fractions were recovered and every other fraction was analyzed by Western blotting. The column was calibrated with the molecular mass marker proteins thyroglobulin (669 kDa), ferritin (440 kDa), catalase (232 kDa), and aldolase (158 kDa) (Amersham Pharmacia Biotech).
Co-immunoprecipitation-For in vitro co-immunoprecipitations proteins were translated in rabbit reticulocyte lysate in the presence of [ 35 S]methionine using the in vitro transcription/translation kit from Promega. For immunoprecipitations, 3-10 l of the translated products were mixed and preincubated for 1 h at 4°C in HEPES, pH 7.9, 100 mM KCl, 0.5% Nonidet P-40, 0.5 mM EDTA, 0.5 mM DTT, 0.4 mM Pefabloc, and 1 g/ml leupeptin, pepstatin, and aprotinin. After pre-clearance with protein A-Sepharose for 1 h at 4°C, HA or Flag antibody and fresh protein A-Sepharose were added and incubated for another hour. The samples were extensively washed in immunoprecipitation buffer, resuspended in SDS loading buffer, and analyzed by SDS-PAGE and autoradiography.
Metabolic Labeling and Immunoprecipitation-HeLa cells were labeled for 5 h with 100 Ci/ml [ 35 S]methionine essentially as described previously (32). Cells were lysed in 20 mM Tris, pH 7.5, 150 mM NaCl, 0.5% Nonidet P-40, 1 mM EDTA, 1 mM DTT, 10 mM NaF, 8 mM ␤-glycerophosphate, 0.1 mM orthovanadate, 10% glycerol plus protease inhibitors (0.4 mM Pefabloc, 1 g/ml aprotinin, 1 g/ml leupeptin, and 1 g/ml pepstatin A). Extracts were precleared with protein A-Sepharose for 1 h and immunoprecipitation was carried out overnight at 4°C. Precipitates were washed three times with lysis buffer and boiled for 5 min in SDS loading buffer. The supernatant was applied on a SDS-PAGE and analyzed by autoradiography.
IB Kinase Assay-To assay for IB kinase activity, HeLa cells were transfected with antisense oligonucleotides against IKAP (GB47), IKK␥ (GB58), or control GeneBlocs (GBC) and 48 h later the cells were treated for 5 min with TNF␣ or IL-1␤. Cellular lysis, immunoprecipitation using IKK␣ antibody, and the in vitro kinase assay were performed exactly as described previously (32).

RESULTS
IKK␥ and IKAP Are Part of Different Molecular Complexes-IKK␥ and IKAP have each been suggested to stably associate with IKK␣ and IKK␤ (26,27,30). To analyze the contribution of IKK␥ and IKAP we tested if both proteins are part of the same complex with IKK␣ and IKK␤ in whole cell extracts from HeLa cells fractionated by gel filtration (Fig. 1A). IKK␣, IKK␤, and IKK␥ had an identical elution profile (fractions 12-16), with an apparent molecular mass between 700 and 900 kDa relative to molecular mass markers, as reported earlier (26). In contrast, IKAP was predominantly found in fractions containing proteins with lower apparent molecular mass (fractions 16 -20), while only small amounts of IKAP co-eluted with the IKK complex and trailed into still much larger apparent sizes. In fact, IKAP peak elution roughly coincided with NF-B⅐IB complexes at an apparent molecular mass of 440 -670 kDa (fractions 18 -22). However, no association of IKAP and NF-B⅐IB complexes was observed after affinity purification with a p65 antibody (data not shown).
Despite the shift of IKK␣ peak elution from fraction 14 to fraction 18 in the absence of IKK␥, the elution of IKAP complexes remained unchanged with maximal elution around fraction 18. This strongly suggests that IKAP is not part of a regular complex with IKK␣ and IKK␥.
To compare associations between IKKs and IKAP or IKK␥ directly, we performed co-immunoprecipitations of in vitro FIG. 1. Endogenous IKK␥ and IKAP are found in distinct complexes as judged by gel filtration chromatography. A, gel filtration analysis of subunits of the IKK and NF-B complexes from whole cell extracts of HeLa cells. HeLa extracts were applied to a Superose 6 gel filtration column and the fractions were analyzed by Western blotting for IKK␤, IKK␣, IKK␥, IKAP, p65, IB␣, and IB␤ proteins, respectively. B, gel filtration analysis of IKAP and the IKK complex in 70Z/3 and 1.3E2 cells. Cytoplasmic extracts of mouse 70Z/3 cells and 1.3E2 cells carrying a mutation which leads to a loss of IKK␥ were fractionated by Superose 6 column. Fractions were compared for IKAP (upper panels) and IKK␣ (lower panels) levels by Western blotting. The specific signal for IKAP was confirmed by peptide competition. Gel filtration of marker proteins was done in parallel and their mobility is depicted above. translated epitope-tagged proteins ( Fig. 2A). HA-IKK␥ was mixed with Flag-IKAP. Whereas HA antibody efficiently precipitated HA-IKK␥ and Flag antibody pulled down Flag-IKAP, neither protein was co-precipitated with the other (lanes 1-4). We also added Myc-tagged IKK␣ and IKK␤ to the IKK␥ and IKAP mixture (lanes 5-8). IKK␣ and IKK␤ could be efficiently co-precipitated with HA-IKK␥, but under these conditions no significant association was seen with IKAP after precipitation with either HA or Flag antibodies. Similar results were obtained in transiently transfected 293 cells, where we observed tight association between IKK␣, IKK␤, and IKK␥, but only a very weak association between IKK␣, IKK␤, and IKAP (data not shown). These observations raised the question whether association between cellular IKKs and IKAP could be detected in intact cells. Whole cell extracts of HeLa cells were used for immunoprecipitations and subsequent Western blotting (Fig.  2B). IKAP was specifically precipitated with the IKAP antibody (lane 3), but only very small amounts of IKAP were precipitated with antibodies directed against either IKK␣ or IKK␥ (lanes 5 and 6) which were also seen with the IgG 1 control (lane 4). Furthermore, the IKAP antibody did not co-precipitate IKK␣ (lane 3). Conversely, co-immunoprecipitation of IKK␣ and IKK␥ using identical conditions was observed with either antibody (lanes 5 and 6). We therefore conclude that IKAP is not stably associated with the IKK complex.
Since NF-B activation by the IKK complex is a cytoplasmic process, components which are involved in this process are expected to reside in the cytoplasm. We analyzed the cytoplasmic and nuclear distribution of IKAP, IKK␣, and IKK␥ and p65 in HeLa cells (Fig. 3A). Whereas IKK␣, IKK␥, IKAP, and p65 were all found predominantly in the cytoplasm (lane 1), a considerable amount of IKAP was also present in the nucleus (lane 2). There, two IKAP bands were detected, both of which were specifically competed by the peptide (data not shown). As IKAP has homology to Elp1, a yeast protein that tightly binds to the chromatin-associated hyperphosphorylated elongating  5-8). Immunoprecipitations were carried out using HA antibody (lanes 2 and 5), HA antibody pre-blocked with the specific peptide (lanes 3 and 6), or Flag antibody (lanes 4 and 7) and analyzed by SDS-PAGE and autoradiography. The input before immunoprecipitation is shown in lanes 1 and 8 and the migrations of the proteins are indicated. B, co-immunoprecipitations of cellular IKAP, IKK␣, and IKK␥. Whole cell extracts of HeLa cells (input lane 1) were precipitated with antibodies specific for IKAP, IKK␣, or IKK␥ (lanes 3, 5, and 6, respectively). Immunoprecipitations were controlled by blockage with a peptide specific for the IKAP antibody (lane 2) or a mouse isotype control antibody (lane 5). Eluates were applied to SDS-PAGE and analyzed by Western blotting. IKK␥ was not detectable in the immunoprecipitations using rabbit antibodies, since the protein has exactly the same migration as the heavy chain Igs.  4, and 6) or IKK␥ and IKAP antibodies which were pre-blocked with the specific peptide (lanes 3 and 5). Immunoprecipitates were analyzed by SDS-PAGE and autoradiography. For the precipitation with anti-IKK␣ and IKK␥ antibodies the specific bands corresponding in size to IKK␣, -␤ and -␥ are depicted with arrows. IKAP and associated proteins (stars) are shown to the right (ns: nonspecific). Positions of molecular weight markers are indicated. form of RNA polymerase II (pol II) (35), we determined whether IKAP is also found in the chromatin fraction. The pellet obtained after the extraction of nuclei with a buffer containing 400 mM NaCl, which contains the chromosomal DNA and stably associated proteins, was boiled in SDS loading buffer and analyzed by Western blotting. Hyperphosphorylated pol II was specifically eluted from the chromatin fraction using these conditions. In contrast to pol II, IKAP was not found in that fraction (lane 3).
Next, the cellular proteins that are associated with IKKs and IKAP were analyzed by immunoprecipitation from extracts of [ 35 S]methionine pulse-labeled HeLa cells (Fig. 3B). As expected, IKK␣ as well as IKK␥ specifically precipitated a complex consisting of IKK␣, -␤, and -␥ (lanes 1-4). No band corresponding to the size of IKAP was detected and no further specific signals were obtained in the range between 25 and 200 kDa. Using an IKAP antibody we specifically co-immunoprecipitated four proteins (lane 6) with sizes of approximately 100, 70, 45, and 39 kDa, none of which co-migrated with the IKKs. These data support that IKKs and IKAP are not stably associated, but that IKAP is part of a different cellular complex with multiple subunits.
Whereas the analysis of the composition of the IKK complex showed that IKAP is not an integral member, it might still function as a potent regulator of IKK activation in response to upstream signaling events. Therefore, we determined the effects of IKAP and IKK␥ on NF-B activation through either overexpression or antisense-mediated down-regulation.

IKAP and IKK␥ Affect NF-B Reporter Activity by Different Mechanisms-Overexpression of IKK␥ or IKAP each have been
shown to repress cytokine-mediated activation of a NF-B-dependent reporter, presumably either by disruption of the IKK complex or by competition for upstream activators of the IKK complex (28,30). To compare the effects of IKAP and IKK␥, either molecule was co-transfected together with an NF-B-dependent ELAM promoter-luciferase construct (ELAMluc) (Fig.  4A). A thymidine kinase promoter-luciferase reporter (TKluc) was used as internal control. Both, IKAP and IKK␥ repressed TNF␣-induced activation of NF-B in the reporter assay (Fig.  4A, upper and middle panel). To test whether the inhibition by IKK␥ or IKAP was restricted to cytokine signaling or whether it could also be observed upon overexpression of an IB kinase, we stimulated NF-B by overexpression of HA-IKK␤ (Fig. 4B). Whereas IKAP inhibited IKK␤-mediated NF-B activation, IKK␥ expression hyper-activated the NF-B reporter, at least at low concentrations. Strikingly, we observed that the effect of IKAP overexpression was not solely restricted to the NF-B reporter, but also resulted in a marked reduction of TK promoter activity (Fig. 4, A and B, lower panel). In contrast, IKK␥ had no effect on TK promoter activity at concentrations where it efficiently blocked the NF-B-dependent promoter. Furthermore, IKAP somewhat repressed HA-IKK␤ expression, the transcription of which is driven by an ␤-actin promoter (Fig.  4B, bottom). To determine whether IKAP had a general effect on promoter activity, we tested its effect on cytomegalovirus and Rous sarcoma virus promoter constructs (Fig. 4C). IKAP repressed both reporter constructs to a similar extent while IKK␥ had no significant effect. Thus, IKK␥ specifically represses NF-B activation probably by sequestration of upstream activators of the IKK complex. In contrast, repression by IKAP is not restricted to NF-B-driven promoters but seems to affect promoter activity in general.
Down-regulation of Cellular IKK␥, but Not of IKAP, Inhibits NF-B Activation-To functionally compare the requirement of cellular IKAP and IKK␥ for cytokine-mediated NF-B activation we reduced the expression levels of both proteins by transient transfection of antisense oligonucleotides (GeneBlocs) in HeLa cells. The effects of the GeneBlocs were first analyzed by RT-PCR. Efficient and specific reduction of the respective mRNAs was observed (Fig. 5A). Next, we tested how downregulation of IKK␥ and IKAP mRNAs affected the respective protein levels and the induction of NF-B by TNF␣ (Fig. 5B). Antisense oligonucleotides directed against IKAP (lanes 3-6) significantly decreased IKAP protein levels. Likewise, transfection of an IKK␥ GeneBloc (lane 7 and 8) led to a reduction of IKK␥ protein levels. The GeneBlocs were specific as they had no effect on expression levels of IKK␣ or p65. We observed that a decrease of IKK␥ protein caused a reduction of NF-B activation in response to TNF␣, as assayed by mobility shift assay (Fig. 5B, upper panel). A similar reduction of IKAP protein did not comparably influence NF-B activation. As a control, Oct-1 DNA binding activity in the same extracts was not affected by any procedure.
To address the effect of reduced IKK␥ or IKAP protein levels on activation of the IKK complex, HeLa cells were transiently transfected with different antisense oligonucleotides and stimulated with either TNF␣ or IL-1␤. We performed an in vitro kinase assay using the immunoprecipitated IKK complex (Fig.  5C, lower panel), which was specifically precipitated by the IKK␣ antibody (compare Fig. 3C). Transfection of antisense oligonucleotides against IKAP or IKK␥ again significantly reduced the steady state amount of each protein but did not change IKK␣ protein levels (upper panel). The kinase activity of the IKK complex was enhanced in response to TNF␣ or IL-1␤ and phosphorylation of IB␣ was dependent on the presence of serines 32 and 36, as expected (lanes 1-5). Whereas reduction of IKAP had no effect on IKK activity (lanes 6 -8), lowering IKK␥ protein levels led to a significant decrease in cytokine inducibility of the IKK complex. We conclude from these experiments that IKAP is not involved to the extent of IKK␥ in cytokine-mediated activation of IKKs and subsequent NF-B activation.
IKK␣, IKK␤, and IKK␥ Are Sufficient to Constitute a Complex Equivalent to the Endogenous IKK Complex-In gel filtration studies the apparent molecular mass of the IKK complex was determined to be 700 -900 kDa (see Fig. 1 and Refs. 26, 27, 30, and 36, and references therein). The cellular IKK complex contains exclusively IKK␣, -␤, and -␥ (Fig. 3) and if composed of an IKK␣/IKK␤ heterodimer and an IKK␥ homodimer should have a molecular mass of about 250 -300 kDa. We were curious to determine whether the difference between the apparent molecular mass in gel filtration analysis and the theoretical molecular weight was caused by further components or if it resulted from structural properties of the IKK complex. We first tested whether an IKK complex with a large apparent molecular weight could be reconstituted in 293 cells simply by overexpression of IKK␣, IKK␤, and IKK␥. 293 cells were transiently transfected and extracts were subjected to gel filtration chromatography (Fig. 6A). Most of the overexpressed HA-IKK␣/␤ eluted in fractions 18 -22 and were not integrated into endogenous IKK complexes, which peaked in fraction 14 (see Fig. 1 and data not shown). Co-expression of HA-IKK␥ led to a complete shift of HA-IKK␣/␤ to fractions corresponding to an apparent molecular mass of 700 to 900 kDa (Fig. 6A, panels a  versus b). Interestingly, the elution peak of HA-IKK␥, when transfected alone, corresponded to more than 600 kDa while co-expression with IKK␣/␤ caused only a slight shift to higher molecular weight fractions (panels b versus c). Probably, IKK␥ forms higher order complexes when expressed alone, but integrates into the IKK complex when IKK␣ and IKK␤ are present. As expected from our previous results, overexpression of Flag-IKAP had no effect on the elution profile of IKK␣ and IKK␤ (panels d and e), supporting that IKAP is not a stable member of the IKK complex.
To exclude that a component in 293 cells, which is in vast excess over the endogenous IKKs, is causing the apparent high molecular weight of IKK␣/␤/␥ we analyzed recombinant puri-fied components (Fig. 6B). Complex formation was analyzed by gel filtration and Western blotting. Purified recombinant IKK␣ peaked in fractions 21 and 22, corresponding to an apparent molecular mass of approximately 440 -470 kDa. Similar to overexpressed IKK␥, purified IKK␥ eluted in fractions 17-18, corresponding to approximately 670 kDa compared to standard proteins. Cross-linking of IKK␥ in fraction 18 with sulfo-EGS (ethylene glycol bis(sulfosuccinimidylsuccinate)) indicated a homotrimeric complex (ϳ150 kDa, data not shown). For the reconstitution of an IKK complex, equal amounts of IKK␣ and IKK␥, as judged by Coomassie staining (left panel), were mixed and incubated on ice for 45 min. Nearly all IKK␣ and IKK␥ now co-migrated in fractions 14 -16, corresponding to an apparent mass of more than 700 kDa. Almost identical results were obtained when IKK␤ and IKK␥ were used (not shown). These data provide evidence that IKK␣ and/or IKK␤ in conjunction with IKK␥ are sufficient to reconstitute a complex with a hydrodynamic property like that of the endogenous IKK complex.

DISCUSSION
The IKK complex is required for activation of NF-B by all physiological inducers tested to date. Recently the two catalytic domains, IKK␣ and IKK␤, which are responsible for inducible IB phosphorylation, have been identified but their tight regulation is depending on the association of further regulatory molecules. A number of different proteins have been reported to be contained in or to co-purify with IKK complexes, including NF-B and IB proteins, a RelA kinase activity, MEKK1, NIK, mitogen-activated protein kinase phosphatase, and two regulatory components, IKK␥ and IKAP (see Refs. 6, 10, 14, and 37-40, for review). However, it has not been determined which of these proteins are substoichiometric or stoichiometric components. Conceptually, the hydrodynamic property of the IKK complex, mimicking a large molecular weight, has been suggestive of a number of IKK associated components.
In this study, we have investigated which proteins are part of the IKK complex and have analyzed the relative contribution of two previously proposed regulatory subunits, IKK␥ and IKAP. An analysis of the endogenous IKK complex revealed that IKK␣, IKK␤, and IKK␥ co-elute upon gel filtration, while p65, IB␣, or IKAP peak-eluted in different fractions. Lack of IKK␥ in a 70Z/3 mutant cell line decreased the mobility of IKK␣, but not that of IKAP. In co-immunoprecipitations from labeled cells IKK␣/␤ associated with IKK␥ as the sole further protein, and vice versa, while IKAP was found associated with other proteins of approximately 100, 70, 45, and 39 kDa. The interaction studies clearly suggest that a regular IKK complex contains exclusively stoichiometric amounts of IKK␣, -␤, and -␥. This was further substantiated by studies employing transfected, in vitro translated or purified, recombinant components: while IKK␣ or -␤ and IKK␥ were efficiently co-immunoprecipitated, no such strong interaction was detected between IKK␣ or -␤  1 and 2), antisense oligonucleotides specific for IKAP (GB47 and GB49; lanes 3-6), or antisense oligonucleotides specific for IKK␥ (GB58; lanes 7 and 8). 48 h after transfection cells were left untreated or stimulated with TNF␣ for 20 min. Whole cell extracts were prepared and NF-B and Oct1 binding activity was determined by electrophoretic mobility shift assay (upper panels). In parallel IKAP, IKK␥, IKK␣, and p65 amounts were determined by Western blotting (lower panels). C, down-regulation of IKK␥ but not IKAP affects the activation of the IKK complex in HeLa cells in response to cytokines. HeLa cells were transiently transfected with control GeneBlocs (GBC; lanes 1-3), antisense oligonucleotides specific for IKAP (GB47; lanes 4 -6), or antisense oligonucleotides specific for IKK␥ (GB58; lanes [7][8][9]. After 48 h cells were left untreated or stimulated with TNF␣ or IL-1␤ for 5 min. The cytoplasmic extracts were analyzed for IKAP, IKK␥, and IKK␣ amounts, as depicted. For each kinase reaction 75 g of extract was used in an immunoprecipitation using monoclonal IKK␣ antibody. After washing the precipitates, kinase reactions were performed in the presence of either 1 g of recombinant IB␣ (lanes 1, 2, 4, 6 -11) or IB␣S32/36A (lanes 3 and 5). Kinase reactions were applied to SDS-PAGE and phosphorylated proteins were visualized by autoradiography. and IKAP or between IKK␥ and IKAP. If IKK␣ or -␤ and IKK␥ were the only integral components, the hydrodynamic behavior of such a complex should be the same as that of cellular IKK complexes. We showed with overexpressed and recombinant proteins that an IKK␣/␤ heterodimer or an IKK␣ homodimer, respectively, was sequestered by IKK␥ into a complex with a gel filtration elution profile similar to that of the endogenous cellular complex. The apparent molecular mass relative to size markers was more than 700 -800 kDa. Although the exact stoichiometry of the IKK␣, -␤, -␥ subunits within the complex remains to be determined, this illustrates that no further components are required to explain the gel filtration elution profile of the complex. Of note, the molecular weight determined relative to globular markers is only accurate for proteins with a globular shape. The observed migration of the IKK␣⅐␤⅐␥ complex suggests that it has an elongated rather than a globular shape.
We found that purified recombinant IKK␥ interacts equally well with IKK␣ and IKK␤, while, in striking contrast, Rothwarf et al. (27) and Mercurio et al. (18) detected a direct interaction of IKK␥ only with IKK␤, but not with IKK␣. However, in agreement with our results, efficient co-immunoprecipitation of IKK␥ and IKK␣ was shown with IKK␤-deficient embryonic fibroblasts (19).
Whereas the data on the composition of the IKK complex strongly suggest that IKAP is not one of the integral members, the possibility remains that it is an essential weakly associated regulator for IKK activation. Therefore, we determined the contribution of IKK␥ and IKAP in functional assays. Overexpression of IKK␥ specifically inhibited NF-B-dependent reporter activity, while IKAP overexpression inhibited all promoters tested, including NF-B-independent reporter genes. IKK␥ overexpression could repress NF-B activation by sequestering upstream activators of the IKK complex (27). However, in conjunction with overexpressed IKK␤, IKK␥ could also stimulate NF-B reporters, presumably by assembly of even more active complexes containing IKK␤ homodimers. Reduction of IKK␥ levels but not of IKAP, by antisense oligonucleotides significantly reduced cytokine induction of IKK kinase activity and of NF-B DNA binding activity. As a possibility, IKAP protein levels could be in vast excess over the components of the IKK complex and only a subpopulation of IKAP, not involved in NF-B signaling, could be affected by the Gene-Blocs. However, a comparison of transfected Flag-IKAP and Flag-IKK␣ with the endogenous proteins, using anti-Flag, anti-IKK␣, and anti-IKAP antibodies indicated that the proteins existed in comparable amounts (data not shown). The repressive effect of IKAP overexpression on cytokineinduced NF-B activation in transient transfection assays . 48 h after transfection cells were lysed and applied to a Superose 6 gel filtration. Fractions were analyzed for the transfected proteins by Western blotting using either HA or Flag antibodies as indicated. B, an in vitro complex consisting of IKK␣ and IKK␥ has an apparent molecular mass of 700 -900 kDa. Equal amounts of recombinant IKK␣ and IKK␥ as judged by Coomassie staining (upper panel) were loaded either alone (panels a and b) or together after 45 min incubation on ice (panel c) onto a Superose 6 gel filtration column. Fractions 11 to 24 were analyzed by Western blotting. Gel filtration analysis of marker proteins was done in parallel and their mobility is depicted above. seems to rely on different or additional mechanisms. The downregulation of several reporter genes suggests that it may directly act on the level of general transcription, either by direct repression or by titrating out essential components. In line with this, IKAP was also found in the nucleus. Moreover, IKAP is homologous to Saccharomyces cerevisiae Elp1/IKI3 and to similar proteins in Schizosaccharomyces pombe and Arabidopsis thaliana (33,31, and 30% identity, respectively). Yeasts and plants do not have conserved NF-B proteins or signaling pathways, indicating a more general function of IKAP homologues. In fact, the S. cerevisiae protein Elp1 has recently been shown to be part of a pol II elongator complex which contains two further stoichiometric components, Elp2 and Elp3, and associates with hyperphosphorylated pol II (35). The yeast Elp3 subunit is a histone acetylase and inactivation of either Elp1 or Elp3 strongly impairs transcription of glucose-induced genes but not transcription of non-induced genes (35,41). Could IKAP be part of a human pol II elongator complex? IKAP was not associated with a salt elution-resistant chromatin fraction that contained hyperphosphorylated pol II. However, the interaction could have been sensitive to the ionic strength and been lost during the preparation procedure. The migration of IKAP in a gel filtration with an apparent size of more than 500 kDa suggests that it might be associated with other proteins and we have identified cellular IKAP-associated proteins of 100, 70, 45, and 39 kDa (Figs. 1 and 3). While the sizes of Elp2 and Elp3 are 90 and 60 kDa, respectively (35,41), no such proteins were precipitated with IKAP. It is noteworthy that S. cerevisiae Elp3 is phylogenetically highly conserved and has homologues in S. pombe, Drosophila melanogaster, A. thaliana, or Caenorhabditis elegans of more than 70% identity over the entire coding regions (41). The human Elp3 homologue is more than 75% identical to the yeast protein and also has a size of 60 kDa. 2 The absence of a 60-kDa protein associated with IKAP thus suggests that IKAP is not part of a strict human homologue of the yeast pol II elongator complex, but may function in a different context in general gene activation mechanisms. Cloning of the four IKAP-associated proteins may elucidate the function of the IKAP complex.
How could IKAP be found associated with the IKK complex and NF-B activation in the first place? First, IKAP purification by IB affinity chromatography and its partial co-elution with the IKK complex upon gel filtration was an intriguing hint for a role as component of the IKK complex. This notion gained support from cDNA cloning of a large novel protein with potential protein-protein interaction motifs. Now all these observations appear coincidental. Likewise consistent with the previously proposed role of IKAP, were experiments showing the interaction of IKAP with three distinct kinases binding to distinct domains of IKAP. However, these data relied entirely on highly overexpressed or purified proteins, a condition that can produce artifactual interaction data through weak but unphysiological binding of proteins. The observed inhibitory effect of IKAP on B-dependent transcription is now shown to lack specificity and would have needed testing of many more control reporter constructs in the first place. Lastly, kinase assays with reconstituted IKAP⅐kinase complexes supported a scaffolding role of IKAP for three distinct kinases. The intriguing interdependence of kinase regulation seen in this experiment may still be relevant irrespective of the now questionable physiology of the used scaffolding protein.
Although the present data show that a regular IKK complex can be reconstituted entirely from the three polypeptides IKK␣, IKK␤, and IKK␥, minor variants of the IKK complex may exist that involve distinct polypeptides. Likewise, weak interactions of the IKK complex with other components may allow for a highly dynamic regulation in addition to the regulation contributed by more static structural components.