IκB Kinase ϵ Interacts with p52 and Promotes Transactivation via p65*

The members of the NF-κB transcription factor family are key regulators of gene expression in the immune response. Different combinations of NF-κB subunits not only diverge in timing to induce transcription but also recognize varying sequences of the NF-κB-binding site of their target genes. The p52 subunit is generated as a result of processing of NF-κB2 p100. Here, we demonstrate that the non-canonical IκB kinase ϵ (IKKϵ) directly interacts with p100. In a transactivation assay, IKKϵ promoted the ability of p52 to transactivate gene expression. This effect was indirect, requiring p65, which was shown to be part of the IKKϵ-p52 complex and to be phosphorylated by IKKϵ. These novel interactions reveal a hitherto unknown function of IKKϵ in the regulation of the alternative NF-κB activation pathway involving p52 and p65.

The members of the NF-B transcription factor family activate defense responses against pathogens and cellular stress. A specific response to a stimulus is achieved through differential activation and dimeric complex formation of the five members of this family: NF-B1 or p105/p50; NF-B2 or p100/p52; and the Rel subfamily, p65 (also termed RelA), RelB, and c-Rel. Upon pathway activation, the IB subunits, which mask the Rel transcription activators, become phosphorylated by the IB kinases (IKKs) 2 and then ubiquitinated and ultimately degraded (1,2). The active NF-B subunits are consequently released, translocate to the nucleus as dimers to bind their cognate target sites, and undergo further phosphorylation, which promotes their ability to transactivate gene expression. p50 and p52 are different from the other factors in that their activation is not controlled by IB subunits but rather by the inhibitory ankyrin repeats in the C termini of their fulllength forms, p105 and p100. These inhibitory domains have to be phosphorylated and processed to produce the respective active p50 and p52 forms (1)(2)(3)(4)(5).
NF-B factors can form homo-or heterodimers, which have different dynamics and DNA target specificity (6,7). The canonical p65/p50 heterodimer is activated by IKK␤ in the IKK␣-IKK␤-NEMO (NF-B essential modulator) complex (8 -12) and binds to its target sequence usually within 0.5 h after stimulation to induce transcription of immediate response genes such as those encoding interleukin (IL)-1␤ and tumor necrosis factor-␣ (TNF␣). Activation of the alternative RelB/ p52 dimers is delayed compared with activation of p65/p50, and the two complexes bind specifically to different variations of the consensus NF-B-binding site (13,14). Some NF-B dimers are transcriptionally inactive and can inhibit the formation of active transcription complexes. For example, increased p52 levels in a human B-cell line has been observed to strongly repress p65-dependent TNF␣ promoter transactivation, hinting at inhibition of p65/p50 by p52 (15). Also, RelB is known to form inactive transcription complexes with p65 (16).
The alternative NF-B pathway, which results in activation of p52, regulates B-cell maturation and is crucial for lymphoid organogenesis (17)(18)(19). It is activated by members of the TNF␣ family such as lymphotoxin-␤ and CD40L (20,21), BAFF (B-cell-activating factor of the TNF family) (22), and TWEAK (TNF-like weak inducer of apoptosis) (23) or upon viral infection with Epstein-Barr virus (24,25) or human T-cell leukemia virus (26). Some stimuli such as CD40L (21,27) and the lymphotoxin-␤ (20) can induce activation of both pathways. Signaling events leading to p100 processing have been subject to intensive studies in recent years. Nuclear shuttling has been linked to regulation of p100 processing (28), which has also been described to be a co-translational event (29). In contrast to the canonical NF-B p65 pathway, p100 does not employ IKK␤ for activation but has instead been shown to be phosphorylated by IKK␣ in an NF-B-inducing kinase (NIK)-dependent manner (30), which is prerequisite for ubiquitination and processing of p100 into p52. Induced and constitutive/pathogenic p100 processing are mechanistically similar. Processing is suppressed by the formation of a presumptive three-dimensional domain consisting of the C terminus, the dimerization domain, and the nuclear translocation domain (31). Inducible processing is crucially controlled by NIK, which activates IKK␣, causing it to phosphorylate p100 (32). Constitutive processing also requires phosphorylation by IKK␣ and is regulated by nuclear shuttling. Different from physiological processing, these pathogenic processing events are NIK-independent (32,33).
The non-canonical IKK⑀ (IKKi) was first described as a lipopolysaccharide-inducible serine/threonine kinase and was hence thought to be essential in the host response against bacterial pathogens (34,35). Expression of IKK⑀ is also induced by proinflammatory cytokines such as TNF␣, IL-1␤, and IL-6, with preferential expression in thymus, spleen, and peripheral blood lymphocytes (34 -36). IKK⑀ is closely related to the constitutive and ubiquitous TBK-1 (TANK-binding kinase-1) (37,38) and shows structural and sequence homology to IKK␣ and IKK␤. Both IKK⑀ and TBK-1 are downstream kinases in Toll-like receptor pathways and are crucial for regulation of interferon (IFN)-␤ and IFN-inducible genes (37, 39 -41). The two kinases share an overlapping substrate specificity for IFN-regulated transcription factors (IRFs), with IKK⑀, preferring IRF-7 over IRF-3 and TBK-1 having the opposite preference (42)(43)(44). In addition, IKK⑀ phosphorylates IB␣ at Ser 36 as well as NF-B p65 at Ser 536 and Ser 468 (34,35,45,46). However, studies have shown that mice bearing a deletion of the IKK⑀ gene respond normally to lipopolysaccharide and double-stranded RNA challenges with respect to activation of IRF-3 and NF-B (39,46). This is presumably due to a functional redundancy between IKK⑀ and TBK-1. IKK⑀ has therefore been linked to the canonical p65/p50 activation pathway and to the Toll-like receptor-3 and -4 pathways to IRFs and IFN-inducible genes (37, 39, 40 -42, 45, 47) but may not have a primary role in their pathways. A defect in CCAAA/enhancer-binding protein-␦-regulated gene expression was noted in lipopolysaccharide-stimulated IKK⑀-deficient cells (40), suggesting a possible link between the NF-B and CCAAA/enhancer-binding protein pathways.
Here, we demonstrate that IKK⑀ can be found in a complex with p52 and p65 and is required for transactivation of gene expression by this complex. We have therefore found a role for IKK⑀ in the activation of an NF-B complex containing p52 and p65.

EXPERIMENTAL PROCEDURES
Cell Culture, Plasmids, and Reagents-HEK293 and HeLa cell lines were purchased from the Centre for Applied Microbiology & Research (Salisbury, UK). Mouse embryonic fibroblasts (MEFs) with a deletion of p65 and the corresponding wild-type MEFs were kindly provided by Ron Hay (University of St. Andrews, Fife, Scotland, UK). The cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal calf serum and incubated at 37°C in a humidified atmosphere of 5% CO 2 . The cells were seeded at 0.5-1 ϫ 10 5 /ml and incubated overnight for experiments. Cells were treated as indicated in the figure legends. Lipopolysaccharide from Escherichia coli serotype O26:B6 and TNF␣ were purchased from Sigma.
Constructs bearing FLAG-tagged IKK⑀ and the construct encoding the kinase activity-impaired mutant IKK⑀(K38A) were kindly provided by Shizuo Akira (Osaka University, Osaka, Japan). The TBK-1-encoding plasmid was a kind gift from Dr. Makato Nakanishi (National Institute for Longevity Sciences, Aichi, Japan). The constructs encoding IKK␣ and IKK␤ was kindly donated by Tularik (San Francisco, CA). The construct comprising hemagglutinin-tagged p65 was provided by GlaxoSmithKline (Stevenage, UK). The plasmids encoding FLAG-tagged p100 (pRSV-p100) and p52 (pRSV-NF-p52) and the glutathione S-transferase (GST) fusion constructs GST-p65-(1-306) and GST-p65-(428 -551) were a kind gift from Neil Perkins (University of Dundee, Scotland). The NF-B-luciferase construct was a gift from Robert Hofmeister (Universitaet Regensburg, Regensburg, Germany), and the IFN-stimulated response element-luciferase construct was purchased from Clontech. The Gal4-p65 transactivation construct and the Gal4-dependent reporter construct were kindly provided by Lienhard Schmitz (Deutsches Krebsforschungszentrum, Heidelberg, Germany). For construction of the p52-Gal4 transactivation fusion protein, we used the pAB vector, which was kindly donated by Oliver Schmidt (University of Giessen, Giessen, Germany). p52 was PCR-amplified with a forward primer containing an EcoRI restriction site (5Ј-AAGAATTCAAGCT-TCACCATGG-3Ј) and a reverse primer containing a BamHI site immediately upstream of the stop codon (5Ј-CGGGATC-CTTCGCGCCCCGCCC-3Ј). The amplified DNA was then digested with EcoRI and BamHI, and the overhangs were filled with Klenow polymerase (New England Biolabs Ltd., Herts, UK) and ligated with the blunted BamHI site of the pAB vector. Correct orientation was confirmed by sequencing. Overexpressed p52-Gal4 fusion protein was detected in immunoblots.
Yeast Two-hybrid Screen-The C-terminal portion of IKK3 (amino acids 541-716) was generated by PCR using Pfu polymerase (Stratagene) and primers 5Ј-CACCCAGCAGATTCA-GTGCTGTTTGG-3Ј (forward) and 5Ј-TCAGACATCAGGA-GGTGCTGG-3Ј (reverse) against a full-length IKK3 clone (GlaxoSmithKline). The resulting PCR product was subcloned into the TOPO-pENTR vector (Invitrogen) using the directional TOPO-pENTR cloning kit (catalog no. K2400-20, Invitrogen) following the manufacturer's instructions. The resulting entry clones were used in Gateway LR Clonase TM recombination reactions (catalog no. 11791-019, Invitrogen) with Gateway-converted pYTH9 and pYTH16 destination vectors to clone in-frame with the Gal4 DNA-binding domain (DBD) for yeast two-hybrid studies as described previously (48). All constructs were confirmed by automated sequencing. Yeast (Saccharomyces cerevisiae) strain Y190 expressing the fusion protein between the Gal4 DBD and the IKK3 C terminus was selected and transformed with a human H9 T-cell cDNA library (GlaxoSmithKline) to give at least a 1-fold representation of the library. Interacting clones were selected with minimal selective dropout medium minus histidine, leucine, and tryptophan and containing 15 mM 3-amino-1,2,4-triazole (catalog no. A8056, Sigma), followed by production of ␤-galactosidase as determined by a freeze-fracture assay. Plasmid DNA was recovered from yeast using the Yeastmaker yeast plasmid isolation kit (catalog no. 630441, Clontech), and the resulting DNA was transformed into E. coli before sequencing.
Mass Spectrometry-Anti-IKK3 antibody was used to immunoprecipitate endogenous IKK3 from HeLa cell extracts (25 ϫ 10 8 ) as described below. Eluted endogenous IKK3 complexes were then run on 4 -12% gels and separated by gel electrophoresis. Gels were stained overnight with Coomassie Blue. Bands were excised and then reduced with dithiothreitol, alkylated with iodoacetamide, and digested with trypsin using a MassPREP workstation (Waters, Manchester, UK). The resulting peptide mixtures were analyzed by liquid chromatographytandem mass spectrometry (MS/MS) using a CapLC and Q-Tof mass spectrometer (Waters) operating in data-dependent MS/MS mode. Targeted liquid chromatography-MS/MS experiments were carried out by creating a list of expected tryptic fragments for selected proteins. Information on the predicted ions was used to direct sampling of particular peptides derived from mixtures in follow-up liquid chromatography-MS/MS analyses. Peptides and proteins were identified by automated searching of all MS/MS spectra against a GlaxoSmithKline non-redundant protein data base.
Gel Filtration-HeLa cells stimulated with 20 ng/ml TNF␣ for 30 h and lysed in high stringency lysis buffer (50 mM HEPES (pH 7.5), 100 mM NaCl, 1 mM EDTA, 10% glycerol, and 0.5% Nonidet P-40) on ice for 15 min. Approximately 8 g of protein from the whole cell extracts was loaded onto a Sephacryl S-300 26/60 gel filtration column (Amersham Biosciences) and eluted with 50 mM NaHPO 4 2Ϫ . 3-ml fractions were collected for immunoblotting or precipitation. If not used immediately, the samples were snap-frozen in liquid nitrogen.
Transfection-based Reporter Gene Assays-MEFs were seeded at 0.2-0.5 ϫ 10 5 /ml in 24-well plates; incubated overnight; and transfected with GeneJuice transfection reagent (Novagen, Madison, WI) according to the manufacturer's instructions with a total amount of 350 -400 g/well containing 150 ng of p-55UAS G Luc and 50 ng of p52-Gal4 fusion construct or of a construct bearing the Gal4 DBD only (used as a negative control). HEK293 cells were seeded at 0.5 ϫ 10 5 /ml in 96-well plates and transfected with a total of 250 ng of DNA containing 100 ng of p-55UAS G Luc or 30 ng of p52-Gal4, Gal4-p65, or the Gal4 DBD. The assays also contained the plasmid DNA of interest, an empty vector as filler DNA, and 20 -50 ng of Renilla reniformis luciferase construct (used as internal control to determine transfection efficiency). For NF-B-luciferase and IFN-stimulated response element-luciferase assays, HEK293 cells were seeded at 1-2 ϫ 10 4 /well in 96-well plates; incubated overnight; and transfected with a total amount of 250 ng of DNA/well comprising 100 ng of reporter construct, the plasmid DNA of interest, 40 ng of R. reniformis luciferase construct, and empty vector as filler DNA. The cells were lysed in passive lysis buffer (Promega Ltd., Southampton, UK) for 15 min. Cell extracts were monitored 24 -36 h post-transfection for firefly luciferase activity following standard protocols. Activities are expressed as -fold activation over unstimulated empty vector controls. The experiments were carried out in triplicate.
Electrophoretic Mobility Shift Assays-Wild-type MEFs were grown in 15-cm dishes and treated as described in the figure legends. Nuclear extracts were prepared as described previously (49). Nuclear extracts were incubated at room temperature for 30 min with 10,000 cpm double-stranded [␥-32 P]ATP and NF-B binding sequence (5Ј-AGTTGAGGGGACTTTC-CCAGGC-3Ј) in hybridization buffer (10 mM Tris-HCl (pH 7.5), 100 mM NaCl, 1 mM EDTA, 5 mM dithiothreitol, 4% glycerol, and 100 g/ml nuclease-free bovine serum albumin) containing 2 g/ml poly(dI-dC) as a nonspecific competitor. Supershift antibodies specific for p65 (catalog no. sc-8008X, Santa Cruz Biotechnology, Inc.) and NF-B2 (clone K-27, catalog no. sc-298, Santa Cruz Biotechnology, Inc.) were added to the nuclear extracts and kept on ice for 1 h prior to hybridization with the oligonucleotide. Protein-DNA complexes were separated on a 5% native polyacrylamide gel, and complex formation was detected by autoradiography.

RESULTS
IKK⑀ Interacts with NF-B2-To gain insight into the function of IKK⑀ and to expand our understanding of the IKK⑀ signaling cascade, a yeast two-hybrid screen for IKK⑀-interacting proteins was performed. A schematic representation of the IKK⑀ protein and the truncation mutant used for yeast two-hybrid screening is shown in Fig. 1A. Using the C-terminal portion of IKK⑀ containing a helix-loop-helix motif (amino acids 541-716) as a bait, we screened an H9 T-cell cDNA library and identified 33 positive clones that were able to activate reporter gene expression. These included IKK⑀-interacting fusion proteins containing portions of TANK (TRAF family memberassociated NF-B activator; I-TRAF (TNF receptor-associated factor)) and NF-B2. The interaction between IKK⑀ and TANK has been reported previously by Nomura et al. (50). More recently, TANK was shown to link IKK⑀ with the classical IKK complex (51). The amino acid sequence of human NF-B2 (p100/p52) is shown in Fig.  1B, and highlighted in boldface is the sequence incorporated within the positive clone from the yeast two-hybrid screen. The N-terminal region of NF-B2 contains the Rel homology region (Fig. 1C). As the association with NF-B2 was novel, additional studies were undertaken to verify the interaction in cells and to investigate the biological relevance in more detail.
Confirmation of an IKK⑀ Interaction with NF-B2 (p52)-Based on the interaction in yeast, the association of IKK⑀ with NF-B2 (p100/p52) was investigated and confirmed to occur endogenously in HeLa cells. IKK⑀ was immunoprecipitated from HeLa cell lysates prepared before and after stimulation with 10 ng/ml TNF␣ for 24 h. Gel-separated proteins were then Western-blotted with an antibody that recognizes the fulllength NF-B2 protein (p100/p52) (Fig. 2). After TNF␣ stimulation, an endogenous interaction could be detected between IKK⑀ and p100 ( Fig. 2A, lane 2). This was not detected in unstimulated cells (Fig. 2A, lane 1) or when immunoprecipitation was performed with non-immune IgG (lane 3). The processed form of p100 (p52) could not be detected because of the IgG heavy chain migrating at approximately the same size upon SDS-PAGE. In unstimulated HeLa cells, p100 was present as shown by immunoprecipitation using anti-NF-B2 antibody (Fig. 2B, lane 1). The levels of p100 were elevated after TNF␣ treatment (Fig. 2B, lane 2). It was also possible to co-immunoprecipitate p100 from HeLa cells and then detect IKK⑀ in the reverse format (Fig. 2C, lane 2). IKK⑀ also interacted with p52, the processed form of NF-B2 (Fig. 2D, lower panel, lanes 2 and  4, upper band), in unstimulated and TNF␣-stimulated HeLa

. IKK⑀ and NF-B2 interact endogenously in HeLa cells.
A and B, lysates were prepared from HeLa cells that were left untreated or stimulated with TNF␣ (10 ng/ml) for 24 h. Endogenous NF-B2 was then immunoprecipitated (IP) with anti-IKK⑀ antibody (A) or anti-p100/p52 antibody (B) and immunoblotted (IB) with anti-NF-B2 (p100) antibody. C, lysates were prepared from HeLa cells that were left untreated or stimulated with TNF␣ (10 ng/ml) for 24 h. Immunoprecipitates were prepared with anti-NF-B2 antibody, and Western blots were probed with anti-IKK⑀ antibody. D, HeLa cells were stimulated with TNF␣ (20 ng/ml) for 24 h or left untreated as indicated and lysed in low stringency lysis buffer. IKK⑀ was precipitated using 3 g of anti-IKK⑀ antibody and analyzed by SDS-PAGE and immunoblotting for the presence of IKK⑀ and p52 as indicated. Hc, heavy chain; Lc, light chain. cells. TNF␣ treatment induced expression of IKK⑀ (Fig. 2D,  upper panel).
IKK⑀-associated Proteins Identified by Immunoprecipitation and Mass Spectrometric Analysis-To complement the yeast two-hybrid approach, endogenous IKK⑀ complexes were purified by immunoprecipitation, and associated proteins were separated by gel electrophoresis and analyzed by electrospray mass spectrometry. Whole cell lysates were prepared from HeLa cells that were left untreated or stimulated with TNF␣ for 24 h, and IKK⑀ complexes were immunoprecipitated. Gel-separated proteins were stained with Coomassie Blue. All bands and the spaces between the bands were excised and digested with trypsin to generate peptide fragments, which were separated by HPLC. Individual HPLC peaks were analyzed by mass spectrometry. NF-B2 was detected only in IKK⑀ precipitates from TNF␣-stimulated cells.
IKK⑀ and p52 Are Components of a 600-kDa Complex-We next determined whether IKK⑀ and NF-B2 are part of a larger multiprotein complex. We therefore applied cell extracts from TNF␣-stimulated HeLa cells to a Sephacryl S-300 gel filtration column and immunoblotted the elution fractions with antibodies against IKK⑀ and NF-B2. Both proteins eluted in the same fractions corresponding to molecular masses of 587,000 to 625,000 Da (Fig. 3A, fractions 33 and 34). To confirm an interaction of the two proteins in this 600-kDa complex, we immunoprecipitated IKK⑀ from the relevant fractions. p52 was found in the precipitated pellet as shown by immunoblotting of the precipitates (Fig. 3B, upper panel, fraction 33; and lower panel, first lane, upper band), suggesting that IKK⑀ and p52 are indeed components of the same multicomponent complex. The results for p100 in this experiment were inconclusive because a nonspecific protein at 100 kDa masked the possible presence of p100 in the precipitate (data not shown).
IKK⑀ and TNF␣ Promote p52 Transactivation-Having identified NF-B2 as a protein that interacts with IKK⑀, we next investigated what the function of this interaction might be. We first asked whether overexpressed IKK⑀ could promote processing of NF-B2 in HEK293 cells. We failed to detect an increased level of the processed form of NF-B2 (p52) in IKK⑀expressing cells compared with control cells (Fig. 4A, third lane), whereas NIK clearly promoted processing of p100 to p52 (fifth lane). IKK␣ and IKK␤ also failed to cause processing. When tested in luciferase assays, all kinases activated the pathways to their respective NF-B or IFN-stimulated response element target sites (Fig. 4B), confirming that the kinases were indeed functional. Also, full-length NF-B2 was not a substrate for IKK⑀ in an in vitro kinase assay (Fig. 4C, lane 1), suggesting that IKK⑀ alone is not sufficient for phosphorylation of NF-B2. IKK⑀ was able to autophosphorylate, confirming its unimpaired kinase activity in our assay.
We next examined whether IKK⑀ could promote transactivation by p52. In a p52 transactivation assay, we cotransfected HEK293 cells, which express a low level of endogenous IKK⑀, with a plasmid encoding full-length p52 fused to the DBD of Gal4 (p52-Gal4) and a Gal4-driven luciferase construct. As shown in Fig. 4D, IKK⑀ promoted p52-driven transactivation up to 6-fold over the control, whereas a kinase-dead form of IKK⑀ (IKK⑀(K38A)), IKK␣, and IKK␤ were without effect.
When cells were stimulated with 10 ng/ml TNF␣, p52-dependent transactivation was increased by 2.5-fold (Fig. 4E). In addition, TNF␣ had a strong co-stimulatory effect with a level of IKK⑀ that had only a marginal effect on its own, causing a 9-fold enhancement over the control (Fig. 4E). No potentiation of the TNF␣ response was detected with IKK⑀(K38A). These data therefore suggest that IKK⑀ can promote transactivation by p52 and that its kinase activity is required for this response.
IKK⑀-induced p52 Transactivation Is p65-dependent-On its own, p52 is unable to transactivate gene expression (52). It has been shown, however, to interact with p65, which might then mediate transactivation. We therefore next tested for p65 involved in the functional link between IKK⑀ and p52. We first tested for a p65-p52 complex in TNF␣-treated cells by carrying out an electrophoretic mobility shift assay. As shown in Fig. 5A, nuclear extracts from MEFs treated with TNF␣ for 30 min bound to the consensus NF-B site, and the complex contained p65 as expected. This can be seen from the effect of anti-p65 antibody, which caused a supershift of the complex (Fig. 5A,  lane 3). No p52 was detected in the complex (Fig. 5A, lane 2). However, a 16-h treatment time resulted in a complex containing p52 and p65 (Fig. 5A, lanes 5 and 6). Anti-p52 antibody inhibited complex formation.
We also tested whether the two proteins interact endogenously. We performed immunoprecipitation studies with untreated and TNF␣-stimulated HeLa cells from which we immunoprecipitated NF-B2 or p65. As shown in Fig. 5B, NF-B2 co-immunoprecipitated with p65 in unstimulated cells (lanes 1 and 2), with TNF␣ treatment increasing the interaction (lanes 4 and 5), whereas IgG controls were negative (lanes 3 and 6). As shown in Fig. 5C, when endogenous IKK⑀ was immunoprecipitated from HeLa cell extracts and blotted for p65, p65 co-immunoprecipitated with IKK⑀ in TNF␣-stimulated cells (lane 3) and, to a lesser extent, in untreated cells (lane 1). This result implies that a complex between IKK⑀, p65, and p52 can be detected endogenously, particularly in TNF␣-treated cells.
Phosphorylation of p65 by IKK⑀ has been demonstrated previously (45). To confirm this, we used GST-p65-(1-306), which FIGURE 4. IKK⑀ does not induce NF-B2 processing but promotes p52-dependent transactivation. A, HEK293 cells were seeded in 10-cm dishes and transfected with 4 g of DNA from plasmid coding for IKK␣, IKK␤, IKK⑀, TBK-1, or NIK as indicated. Whole cell lysates were prepared with low stringency lysis buffer, and extracts were analyzed by SDS-PAGE and subsequently immunoblotted with anti-NF-B2 antibody. The arrows indicate p100 and p52. B, HEK293 cells were seeded in 96-well plates and transfected with the NF-B-luciferase or IFN-stimulated response element (ISRE)-luciferase construct (100 ng) and 50 ng of expression construct for IKK␣, IKK␤, NIK, IKK⑀, or TBK-1 as indicated. Luciferase activity was determined 24 h post-transfection, and the values were normalized for transfection efficiency with R. reniformis luciferase (40 ng). C, GST-p100 (p100) and GST (Ϫ) were provided as substrates for IKK⑀ immunoprecipitated from HEK293 cells transfected with IKK⑀. Immunoprecipitates from untransfected cells were used as a negative control (ctrl). Autophosphorylated IKK⑀ (lanes 1 and 2) is indicated by the arrow. D, p52-dependent transactivation was assayed in HEK293 cells using 100 ng of p-55UAS G Luc and 30 ng of p52-Gal4 or Gal4 DBD (upper panel). The cells were transfected with 50 and 100 ng of plasmid encoding IKK␣, IKK␤, IKK⑀, or IKK⑀(K38A) (IKK⑀KA) or with empty vector only (Ϫ). Luciferase activity was determined 24 h post-transfection, and the values were normalized for transfection efficiency with R. reniformis luciferase. Expression of the kinases in the luciferase assays was tested by Western blotting (lower panel) using anti-FLAG tag antibody (IKK␣ and IKK␤) or anti-IKK⑀ antibody (IKK⑀ and IKK⑀(K38A)). E, HEK293 cells were treated as described above and transfected with 50 ng of plasmid encoding IKK⑀, IKK⑀(K38A), or IKK␣. In addition, the cells were stimulated with 20 ng/ml TNF␣ 6 h post-transfection or left unstimulated (Ϫ) and incubated for another 24 h. Data are the means of three measurements, with error bars representing S.D. The values are representative of at least three independent experiments.
p65 Promotes p52 Transactivation and Is Required for IKK⑀driven p52 Transactivation-Having shown that p52 interacts with p65, we next tested the ability of p65 to promote p52driven transactivation in a p52-Gal4 activation assay. p65 expression increased the activity of the p52-Gal4 transactivation construct by 38-fold over the control (Fig. 6A), suggesting that p65 indeed significantly promotes p52-mediated transactivation.
We next investigated whether p65 mediates the effect of IKK⑀ in the p52 transactivation assay. IKK⑀ was unable to cause p52-driven transactivation in p65 Ϫ/Ϫ MEFs (Fig. 6B, left panel), unlike in wild-type MEFs (right panel). As shown in Fig. 6C, transfection of a plasmid bearing p65 into the p65 Ϫ/Ϫ MEFs strongly enhanced p52 transactivation, and cotransfection of IKK⑀ into the cells doubled this response. We have therefore identified p52 as a protein that interacts with IKK⑀, with IKK⑀ promoting its transactivating activity in a p65-dependent manner.

DISCUSSION
In this study, we have demonstrated that IKK⑀ interacts with the p52 subunit of NF-B and that this interaction promotes transactivation by p52 via p65. Our study has revealed a novel function for IKK⑀ and also demonstrated that p65 can positively influence p52 function.
A variety of substrates for IKK⑀ have been identified to date, including p65 and IB␣ on the canonical NF-B activation pathway and IRFs, which control expression of IFN-dependent genes (34,35,(42)(43)(44)(45)(46). Although IKK⑀ has been implicated in these major regulatory pathways, the deletion of IKK⑀ does not result in a clear phenotype in knock-out mice, hinting at functional redundancy with other kinases (39,46).
In our search for new IKK⑀ substrates and signaling pathways, we discovered a direct interaction of IKK⑀ with the N-terminal (p52-encoding) domain of NF-B2 in a yeast two-hybrid screen. This was the first indication of a role for IKK⑀ in the alternative NF-B activation pathway. We confirmed our initial interaction results by endogenous immunoprecipitation studies using TNF␣-stimulated cells to enhance IKK⑀ expression. With gel filtration experiments, we then revealed that this interaction is part of a 600-kDa multiprotein complex. In a previous study, IKK⑀ was found in a 350-kDa phorbol 12-myristate 13-acetate-inducible complex (35). This smaller IKK⑀ complex has been studied in detail by Chariot et al. (51), who found that TANK links IKK⑀ or TBK-1 with the IKK␣-IKK␤-NEMO complex in phorbol 12-myristate 13-acetate-stimulated cells. We speculate that the difference in IKK⑀ complex size is due to the different stimuli and duration of stimulation. A kinase-dead mutant of IKK⑀ (IKK⑀(K38A)) efficiently blocks phorbol 12-myristate 13-acetate-induced signal transduction to NF-B, whereas it fails to interfere with TNF␣-stimulated NF-B activation, pointing at further subtle differences in TNF␣-and phorbol 12-myristate 13-acetate-activated signaling (35). Interestingly both IKK⑀ and its inactive mutant promote interaction FIGURE 5. p52 and p65 are part of an IKK⑀-dependent NF-B-binding complex. A, nuclear extracts were prepared from wild-type MEFs, which were stimulated with 20 ng/ml TNF␣ and analyzed in an electrophoretic mobility shift assay using an NF-B site oligonucleotide. Nuclear extracts were preincubated for 1 h with antibodies against p65 (2.5 g) and p52 (2.5 g) prior to addition of the labeled NF-B oligonucleotide, and protein-DNA complexes were separated by 5% native PAGE. B, HeLa cells were stimulated with TNF␣ (20 ng/ml) for 24 h or left untreated as indicated and lysed in low stringency lysis buffer. NF-B2 or p65 was immunoprecipitated (IP), analyzed by SDS-PAGE, and immunoblotted (IB) as indicated. C, immunoprecipitates were prepared with anti-IKK⑀ or anti-p65 antibody, and Western blots were probed with anti-IKK⑀ or anti-p65, respectively. D, GST-p65-(1-306), GST-p65-(428 -551), and GST were provided as substrates for IKK⑀ immunoprecipitated from HEK293 cells transfected with IKK⑀ and stimulated with TNF␣ for 24 h as indicated. Autophosphorylated IKK⑀ and phospho-p65-(428 -551) are indicated by the arrows. Hc, heavy chain.
between TANK and NEMO (51). This suggests that its kinase activity is not the only function of IKK⑀.
To further elucidate the functional relevance of the IKK⑀/ NF-B2 interaction, we investigated the effect that IKK⑀ has on processing, phosphorylation, and transactivation of NF-B2. Induced and constitutive processing of NF-B2 have been characterized in great detail recently; both processing events have been found to depend on phosphorylation of several serine residues in the C-and N-terminal domains of p100 by IKK␣ (31)(32)(33). We have demonstrated that overexpressed IKK⑀ is not sufficient to promote processing of NF-B2. Consistent with previous findings (30), NIK efficiently promotes p100 processing to p52, whereas overexpression of IKK␣ on its own does not affect processing. The lack of effect of IKK␣ is probably due to an absence of NIK in the cells. However, this may also apply to IKK⑀, and we are currently investigating this possibility.
IKK⑀ is a kinase for NF-B p65. It phosphorylates p65 at Ser 536 upon IL-1 stimulation (45) and at Ser 468 upon T-cell co-stimulation (53). Both serines are crucial for p65 transactivation. We performed in vitro kinase assays to test whether NF-B2 is also a substrate for IKK⑀ phosphorylation. Despite a substantial IKK⑀ autophosphorylation, which demonstrated that the kinase was indeed functional, we could not detect any phosphorylation of NF-B2, suggesting that it is not a direct substrate.
NF-B dimers have to be activated through transactivating phosphorylation to induce transcription of their target genes. In contrast to Rel family members, which comprise specific transactivation domains, p52 and p50 achieve transactivating properties through binding of a transactivating partner, such as p65 or Bcl-3, in a ternary complex (52,54). We found that IKK⑀ could promote p52-dependent transactivation. Only the kinase-active form of IKK⑀ had this effect. The kinase function of IKK⑀ is thus crucial for complete transactivation of the p52-DNA binding complex. These results suggest that IKK⑀ phosphorylates the transactivating partner of p52. Because it does not phosphorylate p52, we tested whether p65 is this transactivating partner. We found both p52 and p65 in the same TNF␣-induced NF-B-binding complex and confirmed an endogenous TNF␣-enhanced interaction between these proteins in immunoprecipitation experiments. Notably, we could boost p52-dependent transactivation by Ͼ30-fold when we cotransfected p65 with the p52 transactivation construct. We furthermore showed that IKK⑀ could induce p52dependent transactivation only in the presence of p65 because IKK⑀ was without effect in p65-deficient cells. In addition, IKK⑀ significantly enhanced the effect of p65 on p52 transactivation. These observations strongly suggest that p65 is the IKK⑀-activated transactivating partner of p52. The direct interactions between endogenous IKK⑀, NF-B2, and p65 found in HeLa cells provided additional evidence for this FIGURE 6. p65 potently drives p52-dependent transactivation. A, HEK293 cells were transfected with 100 ng of p-55UAS G Luc, 30 ng of p52-Gal4 or Gal4 DBD, and 5 and 10 ng of a construct encoding p65. B, 100 and 150 ng of a plasmid encoding IKK⑀ were tested in a p52 transactivation assay containing 150 ng of p-55UAS G Luc and 50 ng of p52-Gal4 or Gal4 DBD in wild-type (wt) and p65 Ϫ/Ϫ MEFs (upper panels). The expression levels of IKK⑀ in these cells are shown in the Western blots (lower panels). C, the cells were also transfected with 10 ng of plasmid encoding p65 with or without 150 ng of IKK⑀ plasmid. In all experiments, cells were transfected for 24 h. Unstimulated cells were transfected with empty vector as a control (Ϫ). The measured luciferase reporter gene activity values are normalized for transfection efficiency with R. reniformis luciferase. Data are the means of triplicate determinations, with S.D. shown as error bars. The results shown are representative of three independent experiments. ternary transactivation complex. p65 has been shown previously to be a substrate for IKK⑀ in IL-1 and T-cell receptor signaling (45,53), and we confirmed it as an IKK⑀ substrate upon 24 h of TNF␣ stimulation. In this context, it is worth noting that, similar to p52, p65 is also required in the development of lymphoid organs (55).
We have therefore, for the first time, linked IKK⑀ to the alternative NF-B activation pathway through a direct interaction with NF-B2. This interaction is functionally important because it promotes transactivation of p52-dependent genes in a ternary complex with p65.