The Beginning of the End: IκB Kinase (IKK) and NF-κB Activation*

NF-kB/Rel proteins are dimeric, sequence-specific transcription factors that control a variety of important biological decisions from formation of dorsal-ventral polarity in insects to activation of inflammatory and innate immune responses (reviewed in Ref. 1). NF-kB proteins are related through the Rel homology domain (RHD), which subjects them to a particular type of regulation, centered around nuclear-cytoplasmic shuttling (reviewed in Ref. 2). The RHD serves several functions. It is the dimerization and DNA binding domain, and we have learned in atomic detail how RHDs dimerize and interact with DNA (3). In addition, the RHD contains a nuclear localization sequence (NLS), and most importantly it is the site for binding of inhibitors of NF-kB, the IkBs (reviewed in Ref. 2). The IkBs also form a small family with a core composed of six or more ankyrin repeats, an N-terminal regulatory domain, and a C-terminal domain that contains a PEST motif (reviewed in Ref. 2). By binding to NF-kB dimers, the IkBs mask their NLS and cause their cytoplasmic retention. Some IkBs, such as IkBa, contain a nuclear export sequence and when combining with NF-kB dimers in the nucleus (which the IkBs can presumably enter by diffusion) cause their exportin-mediated transport to the cytoplasm (4). Recently the three-dimensional structures of NF-kBzIkB ternary complexes (composed of the RHDs of p50 and p65 and the ankyrin repeat core of IkBa) were solved (5, 6). These fascinating structures indicate that the ankyrin repeats of IkBa form a slightly bent cylinder through a stacked arrangement of a-helices that compose their ankyrin repeats. The peptide loops that connect these helices make specific contacts with the two RHDs, whose N-terminal Ig-like repeats flank the IkB core; the C-terminal Iglike repeats (responsible for dimerization) contact each other with the IkB cylinder lying on top of them. Although the structures solved by two independent groups differ on the way by which IkB masks the NLS located next to the C-terminal Ig-like repeats of the RHDs (5, 6) it is likely that the first two ankyrin repeats sterically hinder the binding of importins to the NLS of NF-kB.

also recognized that upon cell stimulation with these inducers, NF-B dimers translocate from the cytoplasm to the nucleus where they bind target genes and regulate their transcription. Subsequently, the nuclear translocation of NF-B was found to parallel and depend on induced degradation of IBs (reviewed in Refs. 2 and 7).
Potent NF-B activators can induce almost complete degradation of IBs (especially IB␣) within minutes. This process, which is mediated by the 26 S proteasome (8,9), depends on phosphorylation of two conserved serines (Ser-32 and Ser-36 in IB␣) in the N-terminal regulatory domain of IB (10 -12). Homologous serines are also required for degradation of the Drosophila IB homolog, Cactus (13). Even the substitution of a single serine can considerably inhibit IB degradation. Furthermore, these serines cannot be replaced by threonine, indicating that the kinase that phosphorylates them is serine-specific (12). In the presence of proteasome inhibitors, N-terminally phosphorylated IB␣ accumulates very rapidly, indicating that its phosphorylation precedes its degradation and does not result in dissociation from NF-B (8,9). Phosphorylated IBs undergo without delay a second post-translational modification, polyubiquitination. The major acceptor sites for ubiquitin in IB␣ are arginines 21 and 22, whose substitution with lysines considerably retards its degradation (12,14).
Polyubiquitination involves a cascade of enzymatic reactions, the first of which is ATP-dependent and catalyzed by E1 ubiquitin-activating enzyme to form an E1-ubiquitin thioester. The second reaction is catalyzed by the E2 ubiquitin-conjugating enzymes, which receive activated ubiquitin from E1. The last step in the cascade, the transfer of activated ubiquitin from the E2-ubiquitin intermediate to the substrate, is catalyzed by a third group of enzymes, the E3 ubiquitinprotein ligases (15). The E3 group is very heterogeneous, and most of its members are poorly characterized. Recently, a cell-free system that catalyzes the ubiquitination of N-terminally phosphorylated IB␣ was established and used to show that the only regulated step in the IB degradation pathway is the phosphorylation reaction (16). By contrast, the ubiquitinating activity that specifically recognizes phosphorylated IB is constitutive. Most importantly, Yaron et al. (17) have elegantly employed this cell-free system and cutting edge protein purification and sequence determination technology to molecularly identify the recognition component of the phospho-IBspecific E3 activity. This protein, named E3RS IB , is a member of the F-box/WD-repeat family (reviewed in Ref. 18). Interestingly, other members of this family, which contain an F-box and one or two WD or leucine-rich repeats, are essential components of E3 activities involved in regulated protein degradation (19 -21). In the case of E3RS IB , Cdc4, and Grr1, recognition of the phosphoamino acid embedded within a specific sequence is believed to be mediated by the WD repeats (17,18). The F box, on the other hand, is responsible for binding to Skp1, which in turn binds to members of the Cullin family, such as Cdc53 (20,21). The Cullin subunit of the E3 complex appears to be responsible for recruitment of E2-ubiquitin onto the phosphorylated substrate (18). E3RS IB is identical to ␤-TrCP, which was previously isolated via a two-hybrid screen as a protein that binds to the phosphorylated version of the HIV protein Vpu (22). Phospho-Vpu binds CD4, a T cell membrane protein, to induce its ubiquitination and degradation. Curiously, Vpu contains a sequence very similar to the one surrounding the phosphoacceptor sites of IBs (Fig. 1). The same sequence is also present in ␤-catenin, another protein whose abundance is regulated via a ubiquitin-dependent degradation pathway (23). Furthermore, genetic analysis has shown that degradation of the Drosophila ␤-catenin homolog Armadillo depends on a homolog of E3RS IB called Slimb (24). Thus, rather than serving as a recognition sequence for the IB kinase, the conserved sequence surrounding the IB N-terminal phosphoacceptor sites is a recognition site for E3RS IB , whose binding to IB is strictly dependent on phosphorylation of these sites (17). Indeed, the sequence similarity between the IB and ␤-catenin phosphorylation sites led other investigators to examine and confirm the involvement of ␤-TrCP in IB ubiquitination and degradation (25). Coimmunoprecipitation experiments show that like other F box proteins, E3RS IB also associates with Skp1 and Cul1 (25). However, it remains to be established whether Cul1 rather than other Cullins is a physiological component of the E3 IB complex. In addition, the particular E2 that works in conjunction with E3 IB in vivo needs to be identified.

IB Kinase and Its Regulation
The enzymes that catalyze the ubiquitination of phospho-IB are constitutively active. Therefore the regulated step that dictates the fate of IB is in most cases phosphorylation of the two N-terminal serines. As the E3 IB complex may also be involved in degradation of CD4 and ␤-catenin, the phosphorylation step is also the one responsible for specificity in this pathway. There are only two exceptions to this universal pathway for NF-B activation. The first is activation of NF-B in response to UV radiation, which although dependent on IB degradation does not involve IB phosphorylation at the Nterminal sites (26,27). The second exception is anoxia, which stimulates phosphorylation of IB␣ at tyrosine 42 (28). The tyrosinephosphorylated IB␣ was suggested to bind to the SH2 domain of phosphatidylinositol 3-kinase, which yanks it away from NF-B (29). Tyrosine 42, however, is not conserved in other IBs, and therefore the universality of this pathway is questionable. The control of IB phosphorylation in response to all other NF-B activating stimuli rests on the shoulders of the IB kinase (IKK) complex.
Once it became clear that the key step in NF-B activation was IB phosphorylation, a search for a stimulus-responsive protein kinase catalyzing this event was initiated. This effort bore fruit when a protein kinase activity that is specific for the N-terminal regulatory serines of IBs was identified (30,31). This activity, named IKK, is serine-specific and responsive to a number of potent NF-B activators, most notably TNF and IL-1, which stimulate its activity with kinetics that match those of IB␣ degradation (30). Furthermore, the extent to which IKK is activated seems to dictate the kinetics of IB degradation. Gel filtration experiments suggest that IKK is a protein complex, and indeed protein purification, microsequencing, and molecular cloning resulted in identification of three IKK polypeptides. Two of these polypeptides, IKK␣ (IKK1) and IKK␤ (IKK1), are catalytic subunits (30 -33), whereas the third polypeptide, IKK␥ (also known as NEMO), is regulatory (34,35).
IKK␣ was also isolated through a two-hybrid screen as a protein that interacts with the mitogen-activated protein kinase kinase kinase (MAP3K), NIK (36). Although in overexpression experiments NIK acts as a potent IKK and NF-B activator (36 -39), recent experiments question its involvement in IKK activation by either TNF or IL-1 (40). Furthermore, interaction between NIK and IKK␣ occurs upon overexpression of the two in mammalian cells but was not detected under physiological conditions. In addition, the IKK␣ subunit, which was proposed to be the preferential target for NIK (41), is not directly involved in IKK activation (42). IKK␣ and IKK␤ have very similar primary structures (52% overall identity) with protein kinase domains at their N terminus, a leucine zipper (LZ), and a helix-loop-helix (HLH) motif at their C-terminal portion (Fig. 2). IKK␥/NEMO does not contain a recognizable catalytic domain but is composed mostly of three large ␣-helical regions, including a LZ (Fig. 2). Biochemical analysis indicates that the predominant form of IKK is an IKK⅐IKK␤ heterodimer associated with either a dimer or trimer of IKK␥ (34). An IKK complex-associated protein (IKAP) has also been described and proposed to be involved in IKK activation (43) but is not a readily detected constituent of the IKK complex; therefore its physiological significance and function are not clear.
Native IKK complexes purified from mammalian cells seem to be assembled from IKK␣⅐IKK␤ heterodimers plus an undetermined number of IKK␥ subunits (34). Yet, cross-linking experiments indicate that in vitro IKK␣ and IKK␤ can form both homo-and heterodimers in a manner that depends on integrity of their LZ motifs (33). When examined as transiently expressed proteins in mammalian cells, IKK␣ and IKK␤ exhibit identical activation kinetics and substrate specificities (31,32). Although highly reproducible, such experiments are misleading because the transiently expressed proteins readily interchange with their endogenous counterparts and thus are incorporated into IKK complexes (32). Thus when epitopetagged IKK␣ is precipitated and its associated IB kinase activity is measured it is not clear whether one measures its activity or those of endogenous IKK␣ or IKK␤ with which it associates. Indeed, transiently expressed catalytically inactive IKK␣ and IKK␤ associate with a substantial amount of cytokine-inducible IB kinase activity (32). Nevertheless, overexpression of catalytically inactive IKK␣ or IKK␤ inhibits NF-B activation in response to TNF, measured by translocation of p65/RelA to the nucleus (32).
The kinase activities of IKK␣ and IKK␤ or their abilities to be activated depend on LZ-mediated dimerization, and LZ mutations that interfere with this process abolish kinase activity (32,33). IKK␣ or IKK␤ activity is also abolished by mutations within the HLH motif (32,33). These mutations, however, do not interfere with dimerization or binding to IKK␥. Rather, the HLH motif interacts with the kinase domain and can stimulate its activity when expressed in trans (42). IKK activation also requires an intact IKK␥ subunit. No IKK or NF-B activity can be elicited in IKK␥/NEMO-deficient cells that are treated with TNF, IL-1, endotoxin, or dsRNA (35). In addition, IKK complexes assembled on a mutant of IKK␥ that lacks its C-terminal LZ are refractory to all of these agonists (34). These results provide a genetic proof for the importance of IKK in NF-B activation and suggest that the Cterminal region of IKK␥ is necessary for recruitment of upstream activators.
Activation of IKK depends on phosphorylation of its IKK␤ subunit (42). The first evidence for the role of phosphorylation was obtained by treatment of purified, activated IKK complex with protein phosphatase 2A, which resulted in its inactivation (30). Furthermore, treatment of cells with protein phosphatase 2A inhibitor results in activation of IKK (and NF-B). More recently, incubation of cells with TNF was shown to stimulate the phosphorylation of all three IKK subunits (42). IKK␣ and IKK␤ are phosphorylated exclusively at serines. The location of these serines was biochemically mapped for IKK␤, and by conjecture it can be assumed that equivalent sites are phosphorylated on IKK␣. Two of the IKK␤ phosphoacceptors are located in its activation loop (42), a portion of the kinase domain that is involved in phosphorylation-dependent activation of other protein kinases (44). The non-phosphorylated form of the activation loop folds back onto the kinase domain and interferes with entry of ATP and peptide substrates into the catalytic pocket. Phosphorylation moves the activation loop away from the catalytic pocket, thus allowing its interaction with substrates (45). Replacement of the two phosphoacceptor serines (Ser-177 and Ser-181) of IKK␤ with alanines prevents activation, whereas their replacement with phosphomimetic glutamate residues results in con- stitutive activation (31,42). Interestingly, however, replacement of the two equivalent serines (Ser-176 and Ser-180) in IKK␣ abolishes autophosphorylation of this subunit but has no effect on stimulation of total IKK activity by TNF, IL-1, or the upstream kinases MEKK1 and NIK (42). These results, which underscore the biochemical differences between the two catalytic subunits, strongly suggest that IKK is activated as a result of IKK␤ phosphorylation and that IKK␣ phosphorylation, although concurrent with that of IKK␤, is not essential for stimulation of IB kinase activity. In other words, the IKK␤ subunit and not IKK␣ serves as the target for upstream activators involved in proinflammatory signaling that are recruited to the complex via IKK␥.
Phosphorylation is also involved in negative regulation of IKK activity. In addition to the activation loop, IKK␤ is extensively phosphorylated at its C-terminal region, which contains multiple serines (42). Phosphorylation of these sites depends on autokinase activity. Mutagenesis experiments indicate that the C-terminal autophosphorylation sites are involved in shutoff of kinase activity (42). Replacement of 9 or 10 of the C-terminal serines with alanines results in a mutant whose activation lasts four times longer than that of the wild-type enzyme, whereas substitution of the same sites with phosphomimetic glutamic acid residues results in a mutant enzyme that can hardly be activated. Based on these results, a three-state model was proposed to explain the regulation of IKK activity (42). Initially, the inactive IKK complex is not phosphorylated on its catalytic subunits. In response to upstream stimuli, IKK-Ks are activated and recruited to the complex via IKK␥. This results in phosphorylation of IKK␤ and activation of IKK. We presume that initially only a small fraction of IKK is activated through direct phosphorylation by IKK-Ks. However, through intramolecular trans-autophosphorylation the activated IKK␤ subunit can phosphorylate the adjacent subunit, which can be either IKK␣ or IKK␤ (in the case of a homodimer), as well as other inactive IKK complexes through an intermolecular reaction. Indeed, the mere overexpression of IKK␤ in Sf9 cells is sufficient for its activation, which depends on autophosphorylation at the activation loop. The activated IKK complexes phosphorylate the IB subunits of NF-B⅐IB complexes, triggering their ubiquitin-dependent degradation and activation of NF-B. Concurrently, the activated IKK␤ subunits (and presumably the IKK␣ subunits as well) undergo C-terminal autophosphorylation. This reaction, which is unlikely to be processive, operates as a timing device such that when at least nine of the C-terminal serines are phosphorylated the enzyme reaches a low activity state. This facilitates inactivation of IKK by phosphatases once the upstream signal has disappeared. As the C-terminal autophosphorylation sites are adjacent to the HLH motif they may exert their negative effect on kinase activity by changing the conformation of this intrinsic activator domain and affecting its interaction with the kinase domain. This mode of regulation explains why IKK is usually activated in a highly transient fashion. Because of the ability of IKK␤ to propagate its active state via autophosphorylation at the activation loop it is important to have an active way to reduce kinase activity and render it sensitive to inactivation by a phosphatase. Without this prolonged IKK activation would result in prolonged NF-B activation followed by increased production of both primary and secondary inflammatory mediators. As these mediators can lead to further NF-B activation (46), there is a genuine risk that in the absence of an efficient way to rapidly terminate both IKK and NF-B activities even a minor proinflammatory insult would result in a major catastrophe, such as septic shock. Interestingly, constitutive IKK activation was recently detected in Hodgkin's disease cells (47). This results in constitutive NF-B activation, which protects these cells from induction of apoptosis by radio-and chemotherapy (48). In fact, elevated NF-B and IKK activity may protect numerous types of tumors from apoptosis-inducing therapies (49). Thus, IKK offers a reasonable target for development of new antitumor drugs.

IKK Function
The presence of two closely related, yet distinct catalytic subunits within the IKK complex is a curiosity that raises a few questions and possibilities. Are IKK␣ and IKK␤ completely redundant and thus have identical functions? Alternatively, each subunit may be responsible for phosphorylation of different substrates and may even be subject to differential regulation. Initially, little differences between IKK␣ and IKK␤ were observed, and it was assumed that the two have identical function. Yet more recent experiments, discussed above, suggested that IKK␤ and not IKK␣ is involved in IKK activation by proinflammatory stimuli (42). It therefore became important to use a genetic approach to examine the relative functions of IKK␣ and IKK␤. This was done through the use of gene targeting technology to generate mouse strains deficient in either catalytic subunit.
The first surprising result produced by these experiments was the phenotype of IKK␣-deficient mice. The complete loss of IKK␣ results in perinatal lethality. IKK␣ Ϫ/Ϫ mice are born alive but die within 30 min (50). Newborn IKK␣ Ϫ/Ϫ mice display rudimentary limbs and tail, a large omphalocele, and severe craniofacial deformity, but most striking is their skin, which is taut, shiny, and completely devoid of wrinkles. Histopathological examination reveals that the mutant mice contain limb bones of almost normal size, but they are hidden under their thickened skin (50). The major problem with the limb bones is syndactyly and absence of phalanges, whereas more proximal elements appear normal. Other notable skeletal abnormalities include a partially split sternum, fused vertebrae, and severe truncation of the skull. Yet, the most dramatic change is in the structure of the epidermis; the mutant epidermis is up to 10-fold thicker than normal, whereas the dermis appears unaltered. Increased thickness of the mutant epidermis is due to hyperproliferation of cells at the basal layer (which is normally one cell thick). In addition, there appears to be a block to keratinocyte differentiation such that instead of having a stratified epidermis the mutant mice are covered by a uniform layer of cells. The mutant epidermis lacks the upper layer of keratinized cells resulting in increased adhesiveness and stickiness. Transverse sections reveal that mutant limbs and tail are actually "glued" back to the body (50). As epidermal thickenings, such as the apical ectodermal ridge, are an important source for morphogens that control skeletal development many of the morphogenetic defects in IKK␣ Ϫ/Ϫ mice could be secondary to a primary defect in epidermal differentiation.
The second surprise was that IKK␣ is not required for IKK activation by proinflammatory stimuli. Upon stimulation of IKK␣ Ϫ/Ϫ embryonic fibroblasts, primary keratinocytes, or liver tissue with IL-1, TNF, or endotoxin, normal IKK activation and IB␣ degradation were observed (50). Despite normal induction of IKK   FIG. 3. A three-step model for regulation of IKK activity by phosphorylation. In non-stimulated cells the two catalytic subunits of IKK (IKK␣ and IKK␤) are associated into a heterodimer via their LZs. In this state the activation loop within the kinase domain (KD) is not phosphorylated, the HLH motif contacts the kinase domain, and the C-terminal phosphorylation cluster is not phosphorylated either. Upon phosphorylation of the activation loop of IKK␤ and trans-autophosphorylation of the second subunit (either IKK␣ or IKK␤, in the case of a homodimer) IKK is activated. Once activated, in addition to phosphorylation of IBs, IKK progressively autophosphorylates at the C-terminal serine cluster. When at least nine of the C-terminal serines are phosphorylated electrostatic repulsion alters the interaction between the activating HLH motif and the kinase domain such that the kinase reaches a low activity state. At this state IKK is more prone to inhibition by phosphatase action. activity and IB degradation, IKK␣ Ϫ/Ϫ fibroblasts exhibit an approximately 50% decrease in total NF-B DNA binding activity. Thus although IKK␣ does not play a primary role in IKK activation, it may still be involved in stimulating the translocation of NF-B to the nucleus or enhancing its DNA binding activity. Regardless of the partial defect in NF-B activation, the IKK complex in IKK␣ Ϫ/Ϫ cells is of normal size and exhibits normal regulation. As none of the currently available knockout mouse mutants that are deficient in any of the known NF-B subunits or components of the IL-1 and TNF signaling pathways exhibit a similar phenotype, it is unlikely that the developmental and morphogenetic defects in IKK␣ Ϫ/Ϫ animals are caused by alterations in NF-B activation. Most likely, IKK␣ regulates the activity of a key determinant of keratinocyte proliferation and differentiation. It is the altered regulation of this putative IKK␣ substrate that leads to the morphogenetic defects in IKK␣ Ϫ/Ϫ mice.
By contrast to IKK␣, the IKK␤ subunit fulfills all expectations. Although its loss also results in embryonic lethality, the phenotype of IKK␤ Ϫ/Ϫ embryos is completely different from that of IKK␣ Ϫ/Ϫ embryos at the same developmental stage. IKK␤ Ϫ/Ϫ embryos die approximately at embryonic day (E) 12.5, and histopathological examination reveals that the cause of death is massive liver apoptosis (51,52). By comparison, IKK␣ Ϫ/Ϫ embryos or neonates have a perfectly normal liver. The massive increase in apoptosis of hepatocytes in IKK␤ Ϫ/Ϫ embryos is strikingly similar to the major pathology observed in RelA Ϫ/Ϫ embryos, which die at E14.5 because of liver degeneration (53). Because mice that are deficient in both p65 (RelA) and p50 (NFB1) also die at E12.5 (54), it appears that the loss of IKK␤ results in a more severe decrease in NF-B activity than the loss of p65 (RelA) alone. RelA expression is needed to protect cells from TNF-induced apoptosis (55)(56)(57)(58). Indeed, mice that lack both RelA and TNF are viable and have normally appearing liver (59). Thus the liver degeneration in RelA Ϫ/Ϫ mice is due to increased TNF-induced hepatocyte apoptosis unopposed by NF-B, and most likely this is the cause of death in IKK␤ Ϫ/Ϫ embryos. Correspondingly, IKK␤-deficient cells are unable to activate IKK or NF-B in response to either TNF or IL-1 (51,52). Thus, unlike IKK␣, IKK␤ is absolutely required for activation of IKK and phosphorylation of IBs.
Although these experiments provide rather definitive evidence for the different and non-overlapping functions of IKK␣ and IKK␤ they generate a new dilemma. Does IKK␣ exert its morphogenetic function as a component of the "classical" IKK complex composed of IKK␣, IKK␤, and IKK␥, or does it also function as a stand-alone kinase or a component of a completely different complex (Fig. 3)? An answer to this difficult question will require extensive biochemical analysis of IKK complexes in keratinocytes, the cell type in which IKK␣ exhibits its unique function. Once the biochemical form of IKK␣ involved in keratinocyte differentiation is identified, it will be possible to determine how its activity is regulated and which of its substrates plays a fate-determining role in these cells. Despite the many remaining questions, it is satisfying to witness the rapid progress in understanding IKK function and regulation, given the rather recent discovery of this important protein kinase.