A Novel Ubiquitin-like Domain in I (cid:1) B Kinase (cid:2) Is Required for Functional Activity of the Kinase*

Activation of NF- (cid:1) B requires two highly related kinases named IKK (cid:3) and IKK (cid:2) that share identity in the nature and positioning of their structural domains. Despite their similarity, the kinases are functionally divergent, and we therefore sought to identify any structural features specific for IKK (cid:3) or IKK (cid:2) . We performed bioinformatics analysis, and we identified a region resembling a ubiquitin-like domain (UBL) that exists only in IKK (cid:2) and that we named the UBL-like domain (ULD). Deletion of the ULD rendered IKK (cid:2) catalytically inactive and unable to induce NF- (cid:1) B activity, and overexpression of only the ULD dose-dependently inhibited tumor necrosis factor- (cid:3) -induced NF- (cid:1) B activity. The ULD could not be functionally replaced within IKK (cid:2) by ubiquitin or the corresponding region of IKK (cid:3) , whereas deletion of the equivalent section of IKK (cid:3) did not affect its catalytic activity against I (cid:1) B (cid:3) or its activation by NF- (cid:1) B-inducing kinase. We identified five residues conserved among the larger family of UBL-containing proteins

NF-B 1 describes a family of structurally and functionally related, ubiquitously expressed transcription factors that play pivotal roles in innate and adaptive immunity, inflammation, development, cell growth, and survival (1,2). A defining feature of most NF-B responses is their inducibility. Thus, NF-B proteins are normally sequestered inactive in the cytoplasm of resting cells through interaction with distinct members of a family of inhibitory proteins named the IBs. Following appropriate stimulation, the IB proteins are degraded, and NF-B migrates to the nucleus where it binds to specific DNA motifs within the promoters of its target genes (1,2).
The most intensely studied NF-B activation pathway is that induced by the pro-inflammatory cytokine tumor necrosis factor (TNF)-␣. In response to TNF stimulation, IB proteins (typified by IB␣) become rapidly phosphorylated on two specific N-terminal serine residues, and this signals their subsequent ubiquitination and proteasomal degradation. Degradation of IB␣ liberates NF-B dimers that most commonly consist of the p65 (RelA) and p50 NF-B subunits, and these p65:p50 heterodimers regulate the expression of a wide range of genes that include those of pro-inflammatory cytokines, leukocyte adhesion molecules, and anti-apoptotic proteins (2,3). Arguably, the most important intermediate signaling event in this pathway is the phosphorylation of IB proteins, and tremendous effort from a number of laboratories has identified and characterized the signal-responsive kinase complex responsible for this crucial step (4). This catalytic activity resides in a complex of proteins named the IB kinase (IKK) complex that consists of three core subunits named IKK␣ (IKK1), IKK␤ (IKK2), and NF-B essential modulator (NEMO) that is also know as IKK␥ (5)(6)(7)(8)(9). A number of elegant genetic studies have clearly demonstrated that phosphorylation of IB␣ in response to TNF␣ signaling is absolutely dependent upon IKK␤ and NEMO (10 -16), and this pathway that results in liberation of mainly p50:p65 heterodimers is now named the "classical" NF-B pathway (17). Activation of the classical pathway underlies the vast majority of the known functions of NF-B in immune and inflammatory responses, cell survival, and development (17).
In contrast to the fundamental role for IKK␤ in the classical pathway, IKK␣ appears for the most part to be functionally redundant (12,13,16). Thus IB␣ degradation and activation of NF-B in response to TNF␣ remains intact in cells derived from IKK␣ Ϫ/Ϫ animals, whereas this is completely ablated in both IKK␤ Ϫ/Ϫ and NEMO Ϫ/Ϫ cells (10 -16). However, evidence exists that IKK␣ plays a role in mediating the classical pathway in response to a subset of inducers including receptor activator of NF-B ligand (RANKL) (18), and it is known to be able to phosphorylate IB␣ in vitro, albeit with a lower relative activity than IKK␤ (19). Recently, IKK␣ has been shown to migrate to the nucleus and to perform an epigenetic function by regulating histone phosphorylation in the vicinity of classical NF-B-dependent genes (20,21). In addition, IKK␣ may play a regulatory role within the IKK complex by trans-phosphorylating IKK␤ (22), and loss of IKK␣ appears to affect the expression of a number of genes activated in response to pro-inflammatory cytokines (23), although the mechanisms responsible for these effects remain unclear. Nevertheless, despite these separate lines of evidence, it remains that the classical NF-B pathway is absolutely dependent upon IKK␤ and that IKK␣ is dispen-sable for IKK␤-dependent IB␣ phosphorylation and subsequent NF-B activation.
In contrast to its lack of function in classical NF-B signaling, IKK␣ is the key regulator of a recently described "noncanonical" NF-B pathway (24 -26). In this pathway IKK␣ specifically phosphorylates the C terminus of the NF-B p100 subunit (also known as NF-B2) inducing its ubiquitination and processing, and in a manner analogous with IB␣ degradation, p100 processing releases its N-terminal portion (p52) as a transcriptionally active heterodimer with RelB (24 -26). This series of events is completely independent of IKK␤ and NEMO and functions in IKK␤ Ϫ/Ϫ and NEMO Ϫ/Ϫ cells (24 -26). A critical component of the noncanonical pathway is the upstream kinase NIK (NF-B-inducing kinase) that phosphorylates and activates IKK␣ (27), and activation of this pathway is absent in aly/aly mice that carry a mutated NIK gene (24,27). Studies of aly/aly mice together with accumulated genetic evidence clearly demonstrate that the noncanonical NF-B pathway is critical for the maturation of B-cells and development of lymphoid organs. Consistent with this, p100 processing only occurs in response to signals from a subset of TNF receptor family members involved in lymphoid organogenesis and B-cell maturation as follows: B-cell activating factor receptor, CD40, and the lymphotoxin-␤ receptor (24,25,28). In turn, the known inducers of this pathway are the B-cell activating factor, CD40L, heterotrimeric LT␣1␤2, and LIGHT that also binds the lymphotoxin-␤ receptor. Only five genes have been definitively identified as targets of the noncanonical pathway: the cytokine B-cell activating factor and the chemokines SLC (CCL21), ELC (CCL19), BLC (CXCL13), and SDF-1␣ (CXCL12) (24). Consistent with the genetically defined function of the pathway, these are all involved in either B-cell maturation or the development and function of lymphoid organs.
It is clear from these studies that although they physically interact within the IKK complex, IKK␣ and IKK␤ perform highly distinct functions. Although it has been suggested that IKK␣ functions in the noncanonical pathway via a separate IKK␣-alone complex, such a complex has yet to be described and molecularly characterized (24). Despite this potentially separate complex however, it remains that the IKKs share remarkable structural similarity, and consequently the precise mechanisms that underlie their divergent functions remain unknown. In this regard both kinases possess N-terminal catalytic domains, centrally located leucine zippers through which they interact, and a helix-loop-helix domain that is critical for their activation (29). In addition, IKK␣ and IKK␤ each contain a C-terminal NEMO-binding domain (NBD), which we have demonstrated to facilitate NEMO interaction with both kinases (30,31). In light of these similarities, we therefore sought to identify any novel domains or sequences within either of the IKKs that might account for their functional divergence.
We describe here the identification and functional characterization of a novel ubiquitin-like domain (ULD) we have identified in IKK␤. Such a domain does not exist in IKK␣. We demonstrate that deletion of or mutations within the ULD profoundly affect the function of IKK␤, whereas loss of the identical segment of IKK␣ does not affect its activity. Furthermore, our findings strongly suggest that the IKK␤ ULD plays a fundamental role in regulating the interaction of the IKK complex with p65. We therefore conclude that the ULD is absolutely critical for the induced activity of IKK␤, and we further hypothesize that this novel IKK␤-specific domain contributes to the functional divergence of the IKKs.

EXPERIMENTAL PROCEDURES
Cell Culture and Reagents-HeLa, HEK293, and COS cells were obtained from ATCC (Manassas, VA) and maintained in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum, 2 mM L-glutamine, penicillin (50 units/ml), and streptomycin (50 g/ml). Mouse anti-FLAG (M2) and anti-FLAG-coupled agarose beads were purchased from Sigma. Mouse anti-Xpress was purchased from Invitrogen; rabbit anti-p65 (SA-171) was from Biomol (Plymouth Meeting, PA), and the horseradish peroxidase-conjugated secondary antibodies against either rabbit or mouse IgG were both from Amersham Biosciences. Anti-IB␣ and anti-glyceraldehyde-3-phosphate dehydrogenase were from Santa Cruz Biotechnology (Santa Cruz, CA), and anti-p100 (NF-B2) was from Upstate Biotechnology, Inc. Recombinant human TNF␣ was purchased from R & D Systems (Minneapolis, MN).
Plasmids and PCR Mutagenesis-Full-length cDNA clones of human IKK␣ and IKK␤ were the generous gifts from Dr. Michael Karin (University of California, San Diego). All subcloning and mutagenesis procedures were performed by PCR using cloned Pfu DNA polymerase (Stratagene, La Jolla, CA). All PCR conditions and primer sequences are available upon request. Wild-type and mutated IKK␤ cDNAs were inserted between the KpnI and NotI restriction sites of pcDNA-3.1-Xpress (Invitrogen), and all IKK␣ cDNAs were inserted into the EcoRI and XhoI sites of the same vector. FLAG-tagged versions of wild-type and mutated IKK␣ and IKK␤ were constructed by subcloning into the EcoRI/XhoI and HindIII/NotI sites within pFLAG-CMV2 (Sigma), respectively. Human p65 in pcDNA3 was described previously as were the GST-p65N and GST-p65C constructs (32). Full-length cDNAs encoding human ubiquitin and NIK were obtained from HeLa cDNA by PCR cloning using Pfu DNA polymerase. Point mutations within the ULD were made using the QuickChange® site-directed mutagenesis kit from Stratagene. To make all of the domain substitution mutations in I⌲⌲␤ and IKK␣, an EcoRV site was inserted into the IKK␤-d.ULD or IKK␣-d.78 mutant across the join. The fragments to be inserted into the kinases were generated by using primers flanked by EcoRV sites, and these were then ligated into the EcoRV-cut IKK constructs.
The GST-ULD fusion protein was constructed by inserting a PCRgenerated fragment encompassing the IKK␤ ULD (encoding amino acids Leu 307 to Met 384 ) between the EcoRI and NotI sites within pGEX-4T1 (Amersham Biosciences). A cDNA encoding human NEMO was obtained as described previously (31). GST-NEMO was constructed by subcloning the full-length cDNA into the EcoRI and XhoI sites of pGEX-4T1. The fusion of GST with the first 90 amino acids of human IB␣ (GST-IB␣-(1-90)) that was used as a substrate in the kinase assays was described previously (31). GST proteins were made in Escherichia coli (BL21) by treating transformed bacteria with 0.4 mM isopropyl-␤-D-thiogalactopyranoside (Sigma) and following the manufacturer's protocol for protein recovery provided with the vector.
Interaction Analysis-For GST pull-down analysis, IKK␣ or IKK␤ in pcDNA3.1 were transcribed in vitro and translated in the presence of [ 35 S]methionine using the TNT-T7 Quick system from Promega (Madison, WI). Labeled proteins (1 l of reticulocyte lysate) were incubated with GST alone, GST-NEMO, or GST-ULD (1 g) in 100 l of TNT (50 mM Tris, pH 7.5, 200 mM NaCl, 1% Triton X-100) containing protease inhibitors (Complete Protease Inhibitor Mixture, Roche Diagnostics) at 4°C for 30 min, and 20 l of a 50% (v/v) slurry of glutathioneagarose beads (Amersham Biosciences) was then added and incubated a further 15 min. Proteins were then precipitated and washed extensively in TNT before addition of sample buffer (20 l). Samples were then separated by SDS-PAGE (10%), and the resulting gels were stained with Coomassie Blue, fixed, and then examined autoradiographically.
For transient transfection studies, 1 ϫ 10 6 COS cells grown in 6-well trays were transfected with 1 g of total DNA using the FuGENE 6 transfection reagent (Roche Diagnostics). All DNA/FuGENE 6 incubations were performed at a ratio of 1 g of DNA per 3 l of FuGENE 6 according to the manufacturer's recommended protocol in Opti-MEM medium (Invitrogen). After 48 h, cells were lysed in 500 l of TNT, and then complexes were immunoprecipitated (IP) by using anti-FLAGcoupled agarose beads. A portion of each lysate taken prior to immunoprecipitation (5%) was retained for analysis (pre-IP). Precipitated proteins were analyzed by immunoblotting using epitope-specific (anti-FLAG or anti-Xpress) antibodies that were visualized using enhanced chemiluminescence (ECL) reagents from Amersham Biosciences.
Luciferase Reporter Assay-Dual luciferase reporter assays were performed essentially as described previously (30,31). Briefly, 2.5 ϫ 10 5 HeLa cells grown on 12-well plates were transiently transfected using FuGENE 6 with the NF-B-dependent reporter construct pBIIx-luc (0.2 g/well) together with the Renilla luciferase vector (0.02 g/well). Total DNA concentration in each experiment (1.0 g/well) was maintained by adding the appropriate empty vector to the DNA mixture. Forty-eight hours after transfection, cells were lysed in passive lysis buffer (Promega), and luciferase activity was measured using the dual luciferase assay kit from Promega. In some experiments the levels of transfected proteins in 20 g of lysates were examined by immunoblotting by using appropriate epitope tag-specific antibodies.
Immune Complex Kinase Assay-For immune complex kinases assays, 1 ϫ 10 6 HeLa cells grown on 6-well plates were transiently transfected with 1 g of the FLAG-tagged version of the kinase constructs using the FuGENE 6 reagent as described above. Forty-eight hours after transfection, the cells were either untreated or treated with TNF␣ (10 ng/ml) and then lysed on ice in 500 l of TNT for 15 min. Protein content in each lysate was determined by using a Bio-Rad protein assay kit (Bio-Rad) and then normalized among the samples. Proteins in lysates were immunoprecipitated using anti-FLAG (M2)coupled agarose beads for 1 h at 4°C, and the precipitates were washed extensively in TNT and then kinase buffer (20 mM HEPES, pH 7.5, 20 mM MgCl 2 , 1 mM EDTA, 2 mM NaF, 2 mM ␤-glycerophosphate, 1 mM dithiothreitol, 10 M ATP). Precipitates were then incubated for 15 min at 37°C in 20 l of kinase buffer containing appropriate GST-fused substrate proteins and 10 Ci of [␥-32 P]ATP (Amersham Biosciences). The substrate was then precipitated using glutathione-agarose (Amersham Biosciences) and washed extensively with TNT. Beads were then suspended in 20 l of sample buffer, and samples were separated by SDS-PAGE (10%). Kinase activity was determined by autoradiography.
Sequence Alignment-All sequence alignments were performed using MacVector software from Accelrys (San Diego, CA).

IKK␤ Contains a Novel Ubiquitin-like Domain-
We performed a series of proteomic data base searches using various web-based profiles, motifs, and protein family analysis programs to identify any novel structural or functional domains in either IKK␣ or IKK␤. As expected, the previously described structural features of both kinases, including their N-terminal catalytic domains, central leucine zipper motifs, and C-terminal helix-loop-helix and NEMO-binding domains, were identified through a combination of separate approaches. Most surprisingly, however, analysis of the complete amino acid sequence of human IKK␤ using the ExPASy (Expert Protein Analysis System) molecular biology server and the PROSITE data base (us.expasy.org/cgi-bin/scanprosite) identified a region of 78 amino acids from Leu 307 to Met 384 corresponding to the Ubiquitin_2 (type 2 ubiquitin-like) profile (PROSITE document PDOC00271; PROSITE accession number PS50053). No such domain was detected in either IKK␣ or the related kinase IKKe/IKKi.
Despite this sequence identity with ubiquitin, we noted that the PROSITE data base classified this region of IKK␤ as a false positive for the Ubiquitin_2 profile, suggesting that it does not belong to the larger family of ubiquitin-like domain-containing proteins per se (us.expasy.org/cgi-bin/nicesite.pl?PS50053). Therefore, we further investigated the IKK␤ sequence using the Profilescan Server (hits.isb-sib.ch/cgi-bin/PFSCAN) maintained by the Swiss Institute for Experimental Cancer Research, and this analysis identified the same region as a strong match with a normalized match score of 9.548 (statistical interpretation of match score significance is available at hits.isb-sib.ch/doc/motif_score.shtml). Subsequent search analysis using the Protein Families (Pfam) data base of Alignments and Hidden Markov Models (www.sanger-.ac.uk/software/pfam/search.shtml) identified a shorter ubiquitin domain (Pfam document: PF00240) between residues 319 and 354 of IKK␤, and this was also recognized, together with the longer PROSITE Ubiquitin_2 profile, as a ubiquitin domain (document IPR000626) using the Interpro search analysis program of the European Bioinformatics Institute (www.ebi.ac.uk/interpro/index.html). Therefore, we conclude from this accumulated bioinformatic evidence that IKK␤ but not IKK␣ contains a domain located between residues 307 and 384 (more specifically between residues 319 and 354) that displays significant sequence identity with ubiquitin. Although this region strongly resembles both ubiquitin and the UBLs present in a family of unrelated proteins (33), the false positive classification from the original PROSITE search led us to name this region the UBL-like domain (ULD) of IKK␤. The position of the IKKb ULD and its sequence alignment with human ubiquitin are shown in Fig. 1, A and B. Similar alignment of the corresponding region of IKKa (Ile 307 -Val 384 ) (Fig. 1C) demonstrates the significantly lower degree of similarity and identity with ubiquitin.
The IKK␤ ULD Is Required for Catalytic Function-Ubiquitin-like domains have been identified in an expanding family of proteins (33), which include the yeast DNA repair enzyme Rad23 and its human homologues HHR23A and -B, the yeast cell cycle control protein Dsk2, and its human homologues hPLIC-1 and -2, the causative gene of autosomal-recessive juvenile parkinsonism named Parkin, and the anti-apoptotic protein Bag-1 (33). Furthermore, some proteins (i.e. p59 OASL) contain two copies of the domain within their open reading frames (34,35). The precise role of the UBL within many proteins remains unknown; however, it has been directly demonstrated to be absolutely critical for the biological function of at least a subset of these proteins (36 -40). We therefore wished to determine whether the ULD was required for functional activity of IKK␤.
The UBLs of these other proteins do not function as targets for ubiquitination nor are they ligated in a ubiquitin-like manner to separate target proteins (33). Consistent with this, the IKK␤ ULD does not contain lysine residues corresponding to Lys 48 or Lys 63 in ubiquitin that are required for ubiquitin chain formation. The only conserved lysine between the IKK␤ ULD and ubiquitin is at position 337 (see asterisk in Fig. 1B); when we mutated this to arginine (K337R), the resulting kinase did not differ from the wild type with respect to basal and induced catalytic activity and the ability to form complexes with IKK␣ and NEMO (data not shown). Therefore, it appears that like the wider family of UBL domain-containing proteins, the ULD does not function to facilitate IKK␤ ubiquitination or ubiquitinlike conjugation with other proteins.
To investigate the function of the ULD, we constructed a deletion mutant of IKK␤ lacking the region between residues Leu 307 and Met 384 (inclusive) which we named IKK␤-d.ULD. As shown in Fig. 2A, IKK␤-d.ULD failed to activate NF-B when transiently overexpressed in HeLa cells, whereas similar levels of overexpressed wild-type IKK␤ induced robust NF-B activity in these cells. Furthermore, overexpression of IKK␤d.ULD dose-dependently reduced TNF␣-induced NF-B activity induced in HeLa cells (Fig. 2B). To determine the effects of deleting the ULD on the catalytic function of IKK␤, we overexpressed FLAG-tagged versions of wild-type IKK␤ and IKK␤d.ULD in HeLa cells, and then following incubation with TNF␣ for a range of times up to 120 min, we performed immunoprecipitation kinase assays using GST-IB␣-(1-90) as a substrate. As shown in Fig. 2C, catalytic activity of the wild-type kinase was rapidly induced by TNF␣, reaching a maximum at 5 min and then returning to basal levels after 30 min (see Fig. 2C, lanes 1-6). Remarkably, and despite being expressed to the same extent as the wild-type kinase, IKK␤-d.ULD exhibited no basal or TNF␣-induced catalytic activity against IB␣ (Fig. 2C,  lanes 7-12). We also failed to detect catalytic activation of IKK␤-d.ULD in HeLa cells following interleukin-1␤ treatment (data not shown). Therefore, these findings demonstrate that deletion of the ULD renders IKK␤ catalytically inactive against IB␣ and refractory to pro-inflammatory cytokine-induced activation.
Several studies (39, 41-43) have established that recombinant versions of only the ubiquitin-like domain of several proteins possess biological activity in cellular overexpression stud-ies and in vitro biochemical assays. To determine whether the IKK␤ ULD alone would similarly affect NF-B activation, we transiently transfected HeLa cells with just the ULD, and we performed luciferase reporter assays following incubation with pro-inflammatory cytokines. It should be noted that despite intense effort to immunoblot both untagged or epitope-tagged (i.e. FLAG, Xpress, or HA) versions of the ULD, we were unable to detect this protein in lysates of transfected HeLa, COS, or HEK293 cells (data not shown). We did not detect any toxicity associated with overexpressing the ULD in any of these cell types. Although we could not visualize the protein, we consistently observed that transfection with the ULD significantly reduced TNF␣- (Fig. 2D) and interleukin-1␤-induced (not shown) NF-B activity in HeLa and HEK293 cells. Taken together, the data presented in Fig. 2 clearly identify the ULD as a critical domain required for functional activity of IKK␤.
The ULD Is Not Required for Assembly of the IKK Complex-Previous studies (8,9,30,31) have described the molecular mechanisms through which IKK␤ interacts with both NEMO and IKK␣. Nonetheless, it remains possible that the ULD plays an as yet unidentified role in maintaining either of these molecular interactions. Therefore, we performed GST pull-down analysis and, as shown in Fig. 3A, IKK␤-d.ULD interacted with GST-NEMO to the same extent as both wild-type IKK␤ and IKK␣ (compare lanes 6, 9, and 12). Furthermore, a protein composed of GST fused with the IKK␤ ULD (GST-ULD) did not associate with either of the IKKs (Fig. 3A, lanes 5, 8, or 11) or NEMO (not shown). Finally, when we co-expressed IKK␣-FLAG together with either wild-type IKK␤ or IKK␤-d.ULD in COS cells, we recovered both IKK␤ proteins from lysates by immunoprecipitation using anti-FLAG (Fig. 3B, lanes 4 and 5).
We conclude from these findings that the ULD does not play a role in maintaining the interactions between IKK␤ and IKK␣ or NEMO and is therefore not required for assembly of the "core" IKK complex.
Neither Ubiquitin Nor the Equivalent Region of IKK␣ Can Functionally Replace the IKK␤ ULD-In light of its sequence similarity with ubiquitin, we sought to determine whether the ULD could be functionally replaced within IKK␤ by ubiquitin as described previously for the yeast DNA repair protein Rad23 (44). We therefore constructed a panel of deletion and substitution mutants (Fig. 4A), and we noted that each of these kinases interacted with NEMO and each other to the same extent as the wild-type IKK, verifying that these mutations do not affect the inter-molecular interactions within IKK complex (data not shown). We first tested the ability of a FLAG-tagged version of IKK␤ in which the ULD was replaced with human ubiquitin (IKK␤-Ub) to activate NF-B-dependent luciferase activity. As shown in Fig. 4B, neither IKK␤-d.ULD nor the ubiquitin-containing mutant (IKK␤-Ub) could activate NF-B in a luciferase reporter assay when compared with wild-type IKK␤. Furthermore, similar to IKK␤-d.ULD, IKK␤-Ub was basally catalytically inactive and was not activated following treatment of transfected HeLa cells with TNF␣ (Fig. 4C, lanes  5 and 6). We next questioned whether the IKK␤ ULD sequence could be functionally interchanged with the equivalent region of IKK␣, and we constructed an IKK␤ mutant containing the 78 residues Ile 307 to Val 384 of IKK␣ in place of the ULD (IKK␤-␣.78; Fig. 4A). Similar to the ubiquitin substitution mutant, IKK␤-␣.78 failed to activate NF-B-driven luciferase activity (Fig. 4B) and was catalytically inactive and refractory to TNF␣induced activation (Fig. 4C, lanes 7 and 8). To determine the effects of deleting the corresponding 78 amino acids in IKK␣ on its catalytic activity, we constructed a mutant that we named IKK␣-d.78 (Fig. 4A). As shown in Fig.  4D and in contrast to the effects observed with the similar IKK␤ mutant, deletion of this entire region had no effect on the ability of FLAG-IKK␣ to phosphorylate IB␣ in response to TNF␣. However, despite its ability to phosphorylate IB␣ in vitro, the major function of IKK␣ in NF-B activation is as the critical kinase necessary for phosphorylating the NF-B2 precursor protein p100 in the noncanonical NF-B pathway (24 -26). Activation of IKK␣ in this pathway is dependent upon NIK and results in phosphorylation-induced proteolytic processing of p100 to p52 (24 -26). We therefore tested the effects of deleting the ULD-corresponding region of IKK␣ on its activation by NIK, and as demonstrated in Fig. 4E, IKK␣-d78 remained capable of mediating NIK-induced p100 processing to p52. Hence this region appears to be dispensable for the known physiological functions of IKK␣.
We conclude from these experiments that an intact ULD is exquisitely required for IKK␤ activity and cannot be functionally substituted with ubiquitin or the corresponding region of IKK␣. In contrast, catalytic activity of IKK␣ against both IB␣ and p100 does not appear to require the presence of the equivalent region of that kinase.  1-6) or IKK␤-d.ULD (lanes 7-12) were treated with 10 ng/ml TNF␣ for the times indicated. Proteins were immunoprecipitated from lysates using anti-FLAG, and then half of each sample was subjected to kinase assay (KA) using GST-IB␣-(1-90) as a substrate (upper panel). The other half of each immunoprecipitate was immunoblotted using anti-FLAG (lower panel). D, HeLa cells were transiently transfected with pBIIx-luc together with either vector alone (1st two bars and lanes) or 0.25, 0.5, or 1 g/ml of the IKK␤ ULD. Forty-eight hours later, cells were either untreated (Ϫ) or incubated with TNF␣ (10 ng/ml; ϩ) for a further 4 h prior to lysis and measurement of luciferase activity.
Mutational Analysis of the IKK␤ ULD-It is possible that the effects of deleting the entire ULD on IKK␤ activity might be due to gross structural disruption resulting from loss of a relatively large portion of the kinase. To address this issue we constructed a panel of five smaller subdomain deletion mutants lacking stretches of between 11 and 17 residues within the ULD. The positions of the subdomains that we named regions A to E are illustrated in Fig. 5A. As shown in Fig. 5B, versions of IKK␤ sequentially lacking regions A, B, C, or D failed to activate NF-B in a luciferase reporter assay. In contrast, the mutant lacking region E of the ULD (Del.E) induced NF-B activity to a similar level as the wild-type kinase (Fig. 5B). Consistent with these data, we found that only the IKK␤-Del.E mutant displayed TNF␣-induced catalytic activity against GST-IB␣, which resembled the activity of wild-type IKK␤ (Fig. 5C, lanes 11 and 12). Moreover, although we observed low levels of catalytic activity with each of the other deletion mutants, only IKK␤-Del.D consistently exhibited detectable inducibility in response to TNF␣ stimulation, although the magnitude of this activity was significantly less than either wildtype IKK␤ or IKK␤-Del.E (Fig. 5C, compare lanes 9 and 10 with  lanes 1, 2, 11, and 12). These findings strongly suggest that maintenance of the overall integrity of the region between Leu 311 and Ala 367 of the ULD corresponding to the subdomains designated A-D in Fig. 5A is absolutely critical for the functional activity of IKK␤.
In a further attempt to identify any potentially critical functional sites within the IKK␤ ULD, we performed sequence alignment analysis to determine whether the domain contained any residues at positions conserved among the family of UBL-containing proteins (33). As shown in Fig. 6A, alignment of the IKK␤ ULD (Leu 311 -Met 384 ) with the UBLs of Bag-1, BAT-3, OASL, HHR23A, HHR23B, as well as human ubiquitin identified a cluster of five residues that were conserved among all of the proteins. These residues in IKK␤ were specifically proline at position 347 (Pro 347 ), glutamine at 351 (Gln 351 ), leucine at 353 (Leu 353 ), glycine at 358 (Gly 358 ), and leucine at position 361 (Leu 361 ). When we extended the sequence alignment to over 20 distinct UBL-containing proteins from species ranging from yeast to human, these same residues were conserved among all proteins analyzed (data not shown). We there-fore surmised that the residues at these positions might play an important role in the function of this domain, and to test this we constructed a panel of single point mutants in which each residue was substituted with alanine. Consistent with our previous observations (Fig. 3), all of these IKK␤ point mutants interacted with NEMO and IKK␣ (data not shown).
To test the effects of these alanine substitutions on the ability of IKK␤ to induce transcriptionally active NF-B, we performed a luciferase reporter assay, and as shown in Fig. 6B, the P347A and L361A mutants induced NF-B activity to the same level as wild-type IKK␤. Similarly, although the Q351A mutant tended to be less active, over the course of multiple experiments, its ability to induce NF-B activity did not significantly vary from that of the wild-type kinase (not shown). In contrast, NF-B activity induced by G358A was consistently reduced when compared with wild-type IKK␤, and more strikingly, the L353A mutant did not induce activity above the basal levels observed in vector-alone transfected control cells (Fig.  6B). We were therefore surprised to find that despite its inability to activate NF-B, L353A exhibited TNF␣-induced catalytic activity against GST-IB␣ (Fig. 6C). Similar catalytic activity was also observed for all of the other alanine mutants including G358A (data not shown). The failure of IKK␤-L353A to activate NF-B (Fig. 6B) led us to question whether phosphorylation by the mutant kinase could lead to IB␣ degradation. We therefore transfected HEK293 cells with wild-type IKK␤, IKK␤-L353A, or IKK␤-d.ULD, and we determined the effects on both basal and TNF␣-induced levels of IB␣. As shown in Fig. 6D, consistent with our in vitro kinase assay, transfection of HEK293 cells with both the wild-type and L353A mutant kinases decreased the amount of basal IB␣ in cells (Fig. 6D,  compare lanes 1, 5, and 13). In contrast, IB␣ levels in IKK␤d.ULD transfected were unchanged compared with control (Fig. 6D, lanes 9 and 13). Furthermore, IB␣ was degraded following TNF␣ treatment in wild-type and L353A-transfected cells with similar kinetics as control cells, whereas in d.ULDtransfected cells, IB␣ degradation was impaired (Fig. 6D,  compare lanes 12 and 16). This is consistent with the lack of catalytic activity we observed for IKK␤-d.ULD (Fig. 2C). Taken together, these findings demonstrate that although IKK␤-L353A is capable of phosphorylating IB␣ and causing its  5, 8, and 11), or GST-NEMO (lanes 6, 9, and 12). GST proteins were incubated with [ 35 S]methionine-labeled, in vitro transcribed and translated IKK␣ (lanes 4 -6), IKK␤ (lanes 7-9), or IKK␤-d.ULD (lanes 10 -12), and the resulting precipitated complexes were separated by SDS-PAGE (10%). Input amounts of 35 S-labeled proteins (lanes 1-3) and interacting proteins recovered following pull-down (lanes 4 -12) were determined autoradiographically (upper panels). The relative amount of each GST protein after pull-down was visualized by Coomassie Blue (CB) staining the gel (lower panel). B, COS cells were transiently transfected with the constructs indicated, and 48 h later complexes were immunoprecipitated from lysates using anti-FLAG. The resulting complexes or portions of each lysate (5%) taken prior to immunoprecipitation (Pre-IP) were immunoblotted (IB) by using either anti-Xpress or anti-FLAG as indicated.
degradation in response to TNF␣, this single point mutation prevents the overexpressed kinase from activating transcriptionally competent NF-B. (45,46) have demonstrated that IKK␤ can phosphorylate the serine residue at position 536 within the C terminus of the NF-B p65 subunit and that this phosphorylation is critical for its transcriptional activity. We therefore surmised that IKK␤-L353A might be unable to phosphorylate Ser 536 in p65, thereby accounting for its failure to induce transcriptionally active NF-B (Fig. 6B) despite its ability to phosphorylate IB␣ (Fig. 6C) and cause its degradation (Fig. 6D). To test this hypothesis, we transfected HeLa cells with either p65 alone or p65 in the presence of wild-type IKK␤, IKK␤-L353A, or dominant negative IKK␤ (K44M), and we performed a luciferase reporter assay. Consistent with our hypothesis, we found that wild-type IKK␤ enhanced p65-induced luciferase activity, whereas transcriptional activity of p65 that was co-transfected with either K44M or L353A was severely impaired (Fig. 7A). We next performed immunoprecipitation kinase assays using either the N terminus (residues 1-313) or C terminus (residues 314 -550) of p65 fused with GST as substrates (32), and as expected, neither wild-type IKK␤ nor IKK␤-L353A phosphorylated GST-p65N (Fig. 7B, lanes 3, 4, 7, and 8). In contrast, immunoprecipitated wild-type IKK␤ phosphorylated the C terminus of p65, and this activity was basally maximal in our assay and was not further increased by TNF␣ stimulation. To our surprise, however, GST-p65C was also phosphorylated by IKK␤-L353A, and this phosphorylation could be increased to maximum following stimulation with TNF␣. Therefore, our data clearly demonstrate that IKK␤-L353A mutant can phosphorylate the C terminus of p65 strongly, suggesting that the inhibition of overexpressed p65 observed in Fig. 7A is not due to defective C-terminal phosphorylation.

IKK␤-L353A and IKK␤-d.ULD but Not Wild-type IKK␤ Form a Complex with the NF-B p65 Subunit-Previous studies
In the course of our phosphorylation assays, we observed that L353A but not the wild-type kinase formed a stable complex with p65. To explore this further, we transiently transfected HEK293 cells with FLAG-tagged IKK␤, IKK␤-L353A, or IKK␤-d.ULD together with p65, and we immunoprecipitated IKK␤-associated complexes using anti-FLAG. As demonstrated in Fig. 7C, p65 was detected in immunoprecipitates associated FIG. 4. The IKK␤ ULD cannot be functionally substituted with ubiquitin or the equivalent region of IKK␣. A, the structures of wild-type IKK␤ and IKK␣ and the various deletion and substitution mutants are shown. B, HeLa cells were transiently transfected with pBIIx-luc together with either vector alone (pFLAG-CMV2: Control), FLAG-tagged IKK␤ (WT), or the mutants indicated, and luciferase activity was measured in lysates 48 h after transfection. Expression levels of each construct were determined by immunoblotting (IB) using anti-FLAG (lower panel). C, HeLa cells were transiently transfected with the FLAG-IKK␤ constructs indicated, and after 48 h, cells were either untreated (Ϫ) or treated (ϩ) for 5 min with 10 ng/ml TNF␣. Proteins were recovered from lysates using anti-FLAG and an immune complex kinase assay was performed using GST-IB␣-(1-90) as a substrate (lanes 1-8). Proteins in identical lysates from simultaneously transfected cells were immunoprecipitated and immunoblotted using anti-FLAG (lanes 9 -12). D, HeLa cells were transiently transfected with the FLAG-IKK␣ constructs indicated and then processed for kinase assay (lanes 1-4) and immunoblotting (lanes 5 and 6) as described in C. E, HEK293 cells were transiently transfected with pFLAG-CMV2 (Control; lanes 1 and 2), FLAG-IKK␣ (WT; lanes 3 and 4), or FLAG-IKK␣ (d.78; lanes 5 and 6) either alone (Ϫ) or together with NIK (ϩ), and then resulting lysates were sequentially immunoblotted with anti-p100 (upper panel) and anti-FLAG (lower panel). The positions of p100, p52, and a nonspecific band (n.s.) are indicated.
with both the L353A and d.ULD IKK␤ mutants, whereas it was not pulled down with the wild-type kinase. These results were recapitulated when we transfected HeLa cells with IKK␤ and the ULD mutants and tested their ability to interact with endogenous p65. Thus, FLAG-tagged IKK␤-L353A and IKK␤d.ULD associated with endogenous p65 (Fig. 7D, lanes 3 and 4) although no such interaction with the wild-type kinase could be detected (lane 2). These findings therefore suggest that the failure of L353A to induce transcriptional activity of NF-B is not due to a lack of catalytic activity against either IB␣ or the C terminus of p65 but may instead be related to its ability to physically interact with p65. DISCUSSION In this study we sought to identify regions of IKK␣ or IKK␤ that are unique for either kinase. By using a series of bioinformatic strategies, we identified a novel ubiquitin-like domain within IKK␤ that was not detected in IKK␣. Deletion and small mutations within the ULD profoundly affected IKK␤ function, demonstrating that it is a critical regulatory domain required for activity of the kinase. Furthermore, our domain-swap analysis (Fig. 4) strongly suggests that the domain is absolutely specific for IKK␤ and dispensable for catalytic function of IKK␣. In this regard versions of IKK␤ that either lacked the entire ULD or had this region replaced with either ubiquitin or the corresponding region of IKK␣ were completely catalytically inactive against IB␣ and were unable to induce NF-B-dependent gene expression. This is entirely consistent with a previous report (47) in which the catalytic domain of IKK␤ (residues 1-301) was fused with the C terminus of IKK␣ (residues 301-745), resulting in catalytic inactivity of that chimera against IB␣. It therefore appears that an intact ULD is absolutely critical for IKK␤ to phosphorylate IB␣. In contrast, loss of the ULD-corresponding region of IKK␣ did not affect its ability to both phosphorylate IB␣ and induce p100 processing in response to NIK. Therefore, we hypothesize that the pres-ence of the ULD in IKK␤ contributes significantly to the functional divergence of the IKKs. In this regard, it is possible that the presence of this domain, resembling the highly evolutionarily conserved ubiquitin protein, underlies the more ancient innate immune and inflammatory functions mediated by IKK␤ via the classical NF-B pathway (17). It is tempting to speculate further that the role of IKK␣ in mediating critical aspects of the adaptive immune response (i.e. lymphoid organogenesis and B-cell maturation) may be a result of evolutionary modification within the ULD, resulting in the functional divergence of the kinases. Clearly, however, a full understanding of the precise function of the ULD will be required before any conclusions can be drawn concerning its role in shaping the distinct biological functions of the IKKs.
We have demonstrated that the ULD is not involved in formation of the core IKK complex composed of IKK␣, IKK␤, and NEMO. Thus we found that deletion of the ULD does not affect the ability of IKK␤ to heterodimerize with IKK␣. Furthermore, consistent with our prior identification of the NBD in the extreme C terminus of both IKKs (30,31), deletion of the ULD did not affect the interaction of IKK␤ with NEMO. Previous workers (48) have suggested that in addition to the NBD, a separate interaction domain for NEMO exists within IKK␤. Our data demonstrate that if such a region exists, it is not the ULD as the interaction with NEMO was clearly unaffected following its deletion. We also found that insertion of either ubiquitin or the corresponding region of IKK␣ into IKK␤, or deletion of the equivalent IKK␣ residues or insertion of the ULD into IKK␣ did not affect IKK heterodimerization or the ability of either IKK␤ or IKK␣ to interact with NEMO (data not shown). These findings therefore lead us to conclude that the ULD is not required for the maintenance of IKK complex architecture. It remains an intriguing possibility that the ULD functions as a protein-protein interaction domain that facilitates the association with unknown IKK␤-specific interacting proteins (see below). We considered the possibility that the ULD might play a role in ubiquitination of IKK␤ or might be required to physically conjugate IKK␤ with unknown target proteins. However, we have failed to detect such modifications throughout the course of our experiments. Furthermore, two lines of evidence suggest that the ULD does not play a role in facilitating ubiquitination or ubiquitin-mediated conjugation of IKK␤. First, we initially identified the ULD by using the PROSITE data base as having similarity with the type 2 UBL (UBIQUITIN_2) profile present in a large family of proteins (33). To date, however, no evidence exists to support a role for this type of domain in ubiquitination nor have such domains been reported to conjugate with target proteins. In contrast, the type 2 domain defines a family of proteins in which the UBL functions as a linear noncleavable insertion that appears to have a primary role in maintaining specific protein-protein interactions (33). The second piece of evidence against the role in ubiquitination is that the IKK␤ ULD does not contain lysines in either of the positions that correspond to Lys 48 and Lys 63 in ubiquitin. Lysines at these positions are required for ubiquitin conjugation and chain formation, yet the only lysine that is conserved between ubiquitin and the ULD is Lys 337 of IKK␤ (Fig. 1B). The corresponding lysine in ubiquitin (Lys 27 ) is not a target for ubiquitin chain formation. Nevertheless, to test the potential importance of this residue, we have constructed a lysine to arginine substitution mutant (K337R) of IKK␤, and we found that this mutation did not affect the basal or induced catalytic function or the complex forming capability of IKK␤ (not shown). We therefore believe that the ULD is not a target for IKK␤ ubiquitination or ubiquitin-like conjugation of IKK␤ with other proteins.
As described above, the type 2 UBL is considered to function as a protein-protein interaction domain. Specifically, the domain has been shown to directly target UBL-containing proteins to the proteasome and to play a role in ubiquitin-independent proteasomal protein degradation. In this regard the UBL domains of proteins including the yeast protein RAD23 and its human orthologues HHR23A and -B, hPLIC-1 and -2, Parkin, and BAG-1 facilitate their direct interaction and stable association with components of the regulatory domain of the 26 S proteasome (33, 36 -40, 42, 43, 49 -53). Furthermore, this proteasomal localization does not lead to degradation of these proteins but enables the proteolysis of ubiquitinated cargo proteins that associate with the UBL-containing carrier. Therefore, at least some UBL-containing proteins appear to function as molecular chaperones that present ubiquitinated cargo proteins to proteasomal ATPases where they are subsequently unfolded and degraded, although leaving the UBL protein intact (33, 36 -40, 42, 43, 49 -53).
This function of certain UBL-containing proteins presented a fascinating hypothesis regarding the potential role of the ULD in IKK␤. Thus we considered that the ULD might facilitate an interaction between the IKK complex and proteasome thereby bringing the kinase, its ubiquitinated substrate (i.e. IB proteins), and the degradation machinery into close context. Furthermore, association with the proteasome may explain our failure to detect overexpressed ULD alone as it might be rapidly degraded via this route. Nevertheless, despite intense effort using a wide range of available reagents, we have been unable to detect any interactions between either endogenous or overexpressed IKK␤ and the proteasome. 2 It therefore appears that the IKK␤ ULD does not function in manner similar to the UBLs of HHR23A and -B, PLIC-1 and -2, Parkin, or Bag-1. This finding is perhaps not completely surprising as the UBL is located in the extreme N terminus of all of the UBL-containing proteins that interact with the proteasome, whereas the ULD is centrally positioned in IKK␤. Furthermore, the proteasomeinteracting proteins also contain ubiquitin-associated domains through which they can interact with ubiquitinated proteins. In addition, none of these proteins are kinases and appear to function primarily as molecular adaptors or chaperones. Finally, our domain swap analysis demonstrated that the function of the ULD could not be replaced with ubiquitin, whereas insertion of ubiquitin into the UBL site in Rad23 has been reported to maintain the function of that protein (44). Thus, it appears that the proteasome-binding capability of ubiquitin is insufficient to maintain the function of IKK␤. It is therefore likely that the ULD represents a distinct type of ubiquitin-like domain that performs a function separate from proteasomal localization.
The possibility therefore remains that the ULD plays a role in maintaining interactions with separate IKK␤-specific proteins. One particular set of candidates for such interacting proteins may be components of the COP9 signalsome that is a distinct protein complex that exhibits similarities to the 26 S proteasome (54). Most intriguingly, catalytic activity specific for IB␣ has been found associated with the COP9 signalsome,  1-4) or wild-type IKK␤ (lanes 5-8) and then incubated in the absence or presence of TNF␣ (10 ng/ml) for 5 min. Following lysis and immunoprecipitation using anti-FLAG, an immune complex kinase assay was performed using either GST-p65C (lanes 1, 2, 5, and 6) or GST-p65N (lanes 3, 4, 7, and 8) as substrates. Samples were then separated by SDS-PAGE (10%) and visualized by autoradiography (upper panel). The gel was stained with Coomassie Blue (CB) to identify the substrate proteins (lower panel), and the relative position of GST-p65N on the autoradiograph is indicated (*). Amounts of IKK␤ in each lane are shown in the middle panel. C, HEK293 cells were transfected with p65 together with either pFLAG-CMV2 (control), wild-type IKK␤, IKK␤-L353A, or IKK␤-d.ULD, and then immunoprecipitation (IP) was performed using anti-FLAG. Resulting precipitated material was immunoblotted (IB) using either anti-p65 or anti-FLAG as shown. A portion (5%) of the original lysate (Pre-IP) was immunoblotted using anti-p65 (middle panel). D, HeLa cells were transfected with the constructs indicated, and then anti-FLAG was used to immunoprecipitate complexes from lysates. Immunoprecipitated samples were immunoblotted using either anti-p65 or anti-FLAG as shown.