The NF-κB Activation Pathway: A Paradigm in Information Transfer from Membrane to Nucleus

Science's STKE  26 Oct 1999:
Vol. 1999, Issue 5, pp. re1
DOI: 10.1126/stke.1999.5.re1


Nuclear factor kappa B (NF-κB)/Rel proteins are dimeric, sequence-specific transcription factors involved in the activation of an exceptionally large number of genes in response to inflammation, viral and bacterial infections, and other stressful situations requiring rapid reprogramming of gene expression. In unstimulated cells, NF-κB is sequestered in an inactive form in the cytoplasm bound to inhibitory IκB proteins. Stimulation leads to the rapid phosphorylation, ubiquitinylation, and ultimately proteolytic degradation of IκB, which frees NF-κB to translocate to the nucleus and activate the transcription of its target genes. The multisubunit IκB kinase (IKK) responsible for the inducible phosphorylation of IκB appears to be the initial point of convergence for most stimuli that activate NF-κB. IKK contains two catalytic subunits, IKKα and IKKβ, both of which phosphorylate IκB at sites phosphorylated in vivo. Gene knockout studies indicate that IKKβ is primarily responsible for the activation of NF-κB in response to proinflammatory stimuli, whereas IKKα is essential for keratinocyte differentiation. The activity of IKK is regulated by phosphorylation. IKK contains a regulatory subunit, IKKγ, which is critical for activation of IKK and is postulated to serve as a recognition site for upstream activators. When phosphorylated, the IKK recognition site on IκBα serves as a specific recognition site for the κ-TrCP-like component of a Skp1-Cullin-F-box-type E3 ubiquitin-protein ligase. A variety of other signaling events, including phosphorylation of NF-κB, phosphorylation of IKK, new synthesis of IκBs, and the processing of NF-κB precursors provide mechanisms of modulating the amount and duration of NF-κB activity.


There are several well-characterized signaling pathways that lead to transcriptional activation. Of these, the NF-κB pathway is distinct in the rapidity of its activation, its unusual mechanism of regulation, and the comparatively high level of understanding we have of the entire pathway: from detailed molecular structures to the physiological function of its components. NF-κB is maintained in an inactive form by sequestration in the cytoplasm through interaction with inhibitory proteins, the IκBs. Proteolytic degradation of IκB immediately precedes and is required for NF-κB nuclear translocation. This irreversible step in the signaling pathway constitutes a commitment to transcriptional activation. The signal is eventually terminated through cytoplasmic resequestration of NF-κB, which depends on IκBα synthesis, a process that requires NF-κB transcriptional activity. NF-κB activation often occurs in situations where rapid and decisive action is required for cell survival, such as during activation of the innate immune response--our first line of defense against bacterial, viral, and fungal infections.

NF-κB was first identified in 1986 as a nuclear factor bound to an enhancer element of the immunoglobulin κ light chain gene and was believed to be specifically expressed in B cells (1). It quickly became apparent that NF-κB is present in virtually every cell type but is retained in the cytoplasm in an inactive form bound to IκBs (2-5). Well over 2000 papers dealing with NF-κB have been published. The purpose of this review is not to enumerate all of the experimental data related to NF-κB activation, but to reconcile many of the inconsistencies into unifying hypotheses and pathways, as well as to explore the intriguing possibilities suggested by recent results, especially those that have arisen since the molecular identification of the IκB kinase (IKK). Consequently, although we attempt to provide sufficient background details for those unfamiliar with NF-κB and the major signaling pathways that control its activation, our intention is to avoid duplication of the many fine reviews that have been published over the years on various aspects of NF-κB composition and function. Moreover, because so many different stimuli activate NF-κB and, in many cases, warrant and have received specialized reviews, this review concentrates on the events and features that are common to the major NF-κB signaling pathways, those that are activated by proinflammatory stimuli and converge on IKK. Fortunately, at this time, it is this "common ground" of the NF-κB signaling landscape that is the most clearly determined and where our understanding appears to be expanding most rapidly.


NF-κB/Rel proteins.

In its active DNA-binding form, NF-κB exists as a heterogeneous collection of dimers composed of various combinations of members of the NF-κB/Rel family (Fig. 1). Five mammalian members of this family have been identified, NF-κB1 (p50 and its precursor p105), NF-κB2 (p52 and its precursor p100), c-Rel, RelA (p65), and RelB (3-5). Although it is not certain if additional NF-κB/Rel genes are contained in mammalian genomes, no new family members have been identified within the past 7 years despite the considerable amount of new sequence information generated over that time period. Three family members have been identified in Drosophila: Dorsal, Dif, and Relish (3). All of these proteins share a highly conserved Rel homology region (RHR), a stretch of about 300 amino acids composed of two immunoglobulin (Ig)-like domains. The RHR is responsible for dimerization, DNA binding, and interaction with the inhibitory IκB proteins. A nuclear localization sequence (NLS) is also contained within the RHR. Various NF-κB dimers exhibit different binding affinities for κB sites bearing the consensus sequence GGGRNNYYCC, where R is purine, Y is pyrimidine, and N is any base (4). The Rel proteins also differ in their capacity to activate transcription; of the mammalian family members, only p65/RelA and c-Rel function as potent transcriptional activators, although RelB is able to activate transcription in certain cell types. Dimers composed solely of Rel proteins that lack transcriptional activation domains, such as p50, are thought to mediate transcriptional repression. Targeted disruption of individual Rel loci has been accomplished in mice (3,6). These "knockout" studies indicate specific roles for each NF-κB protein, although surprisingly only p65/RelA is essential for survival. There appears to be some compensation between other NF-κB members. Heterodimers of p65 and p50 were the first form of NF-κB to be identified and are the most abundant in most cell types. Consequently, even though the term "NF-κB" applies to all dimeric forms of the Rel proteins, it is often used to refer to the p50-p65 complex. Currently, it appears that all NF-κB complexes are regulated in the same manner--primarily through interactions with IκBs, although various IκBs may show preference for certain NF-κB dimers.

Fig. 1.

Schematic representation of members of the mammalian and Drosophila Rel/NF-κB and IκB families of proteins. The numbers of amino acids in each protein are listed on the right; mammalian sequences correspond to the human proteins. The arrows point to the approximate location of the COOH-terminal residues of p50 and p52 (following processing of p105 and p100, respectively; see text). The pairs of serine residues required for inducible degradation of IκBα, IκBβ, IκBϵ, and Cactus are represented by SS; LZ, leucine zipper; GRR, glycine-rich region; SRR, serine-rich region. The number and position of ankyrin repeats were determined on the basis of homology and the x-ray structure of IκBα (17, 18).

IκB proteins.

The IκB family includes IκBα, IκBβ, IκBγ, IκBϵ, Bcl-3, the precursors of NF-κB1 (p105) and NF-κB2 (p100), and the Drosophila protein Cactus (3,7) (Fig. 1). Of these, the most important regulators of mammalian NF-κB are IκBα, IκBβ, and IκBϵ. Alternative splicing results in two isoforms of IκBβ, IκBβ1 and IκBβ2 (8). All IκBs contain either six or seven ankyrin repeats, and these stacked helical domains bind to the RHR and mask the NLS of NF-κB. Only IκBα, IκBβ, and IκBϵ contain NH2-terminal regulatory regions that are required for stimulus-induced degradation, the key step in NF-κB activation. IκBs also have an important role in the termination of NF-κB activation. Newly synthesized IκBα enters the nucleus and binds NF-κB, thereby enhancing its dissociation from the DNA (the affinity of NF-κB for IκB appears to be higher than its affinity for κB sites on DNA) and causing its reexportation to the cytoplasm by means of a nuclear export sequence (NES) present on IκBα (9).

IκBα was the first IκB family member to be cloned and is by far the best characterized. When IκB function is mentioned in the literature, almost invariably it is IκBα that is intended. The physiological properties of the other IκBs are for the most part poorly understood. Mice lacking Bcl-3 and IκBα have been generated (3,6). Bcl-3-/- mice exhibit specific defects in response to certain immunogenic agents. IκBα-/- mice are born normally but die 7 to 10 days after birth and exhibit inflammatory conditions of the skin consistent with increased NF-κB activity (10,11). IκBα-deficient embryonic fibroblasts do not have increased NF-κB activity; however, sustained activation of NF-κB does occur in these cells after stimulation with tumor necrosis factor α (TNF-α) (10,11). IκBα is unique among the IκB proteins in that its expression is controlled by NF-κB. Consequently, although endogenous IκBβ does not compensate for the lack of IκBα in IκBα -/- mice, when placed under control of the IκBα promoter, IκBβ can effectively substitute for IκBα (12).

Structure of NF-κB and IκB.

The butterfly-like x-ray crystal structures of various forms of NF-κB bound to DNA have revealed the structural origins of the transcription factor's sequence and dimerization specificities (13-16). The RHR contains two Ig-like domains connected by a flexible linker region. Both domains contact the DNA; loops within the NH2-terminal domain are primarily responsible for sequence specific recognition, whereas the COOH-terminal domains contain the dimer interface. Dimerization is mediated by extensive hydrophobic interactions along the interface surface, which is formed by a three-stranded β sheet packing against a similar sheet in the opposing molecule. Two independent structure determinations were reported for NF-κB:IκB ternary complexes consisting of the RHR of p65, the COOH-terminal Ig-like domain of p50 (including the NLS), and the ankyrin repeats of IκBα (17-20). The ankyrin repeats of IκBα form a curved stack of α helices that bind to the COOH-terminal dimerization domains of the RHRs in a discontinuous fashion. In both structures, the NES on IκBα is exposed and disruption of NF-κB DNA binding appears to be primarily mediated by residues COOH-terminal to the ankyrin repeats of IκBα. Unfortunately, the data do not unambiguously address the manner in which IκBα masks the NLS of NF-κB. This is due in part to differences between the two structures in the p65 NLS and the lack of sufficient ordered structure in the p50 NLS. Also, both structures lack the critical NH2-terminal regulatory domain of IκBα. Nevertheless, it seems likely that the NH2-terminal ankyrin repeats of IκBα sterically hinder the binding of karyopherins to the NLS of NF-κB.

Biomedical importance of NF-κB.

NF-κB regulates the transcription of an exceptionally large number of genes, particularly many that participate in immune and inflammatory responses (5, 21,22). NF-κB activation also functions in the antiviral response through regulation of interferon gene expression (23,24). Yet, through adaptation, many viruses including human immunodeficiency virus type 1 (HIV-1) (25) and human T-cell leukemia virus type 1 (HTLV-1) (26) that do not cause increased expression of interferon exploit NF-κB to activate their own genes and to stimulate the survival and proliferation of the lymphoid cells in which they replicate. In most cells, NF-κB mediates cell survival signals, protecting cells from apoptosis, but under certain conditions and in certain cell types it may also induce apoptosis (27,28). Inappropriate regulation of NF-κB contributes to a wide range of human disorders, including cancers (29,30), neurodegenerative diseases (31), ataxia-telangiectasia (32), arthritis (33), asthma (34), inflammatory bowel disease (35), and numerous other inflammatory conditions.

Activating signals.

NF-κB is activated by many distinct stimuli, including proinflammatory cytokines such as TNF-α and interleukin 1 (IL-1), T and B cell mitogens, bacteria and bacterial lipopolysaccharide (LPS), viruses, viral proteins, double-stranded (ds) RNA, and physical and chemical stresses (3,5). It is beyond the scope of this review to discuss all of the properties of these activation pathways; many of their common features will be presented in later sections.

Overview of the consensus NF-κB activation pathway.

Although there is no full understanding or agreement as to how various extracellular and intracellular stimuli trigger NF-κB activation or at what point their signaling pathways converge, several common features do exist for all of the major activation pathways studied (Fig. 2). Potent activators, such as TNF-α, IL-1, or LPS, induce the rapid degradation of the IκBs (especially IκBα) within minutes. In the case of IκBα, this degradation process consists of a series of well-characterized steps (2,3). Inducible phosphorylation of IκB occurs at Ser32 and Ser36 in IκBα, and mutation of either serine (even to a threonine) greatly inhibits the degradation process (36-41). Phosphorylation leads to the immediate recognition of the NH2-terminus of IκBα by the recently identified β-TrCP-like F-box-containing component of an Skp1-Cullin-F-box (SCF)--type E3 ubiquitin-protein ligase complex (42-44), which consequently results in the polyubiquitinylation of IκBα primarily at lysines 21 and 22 (39,45-47). This modification then targets IκBα for rapid degradation by the 26S proteasome. The degradation of IκB results in exposure of the NLS of NF-κB, which results in translocation of NF-κB to the nucleus, presumably as a consequence of binding to karyopherins, proteins responsible for nucleocytuplasmic transport within the cell (48,49). Inhibitors of the 26S proteasome efficiently block nuclear translocation of NF-κB, indicating that neither phosphorylation nor ubiquitinylation is sufficient to dissociate IκB from NF-κB (38, 50,51).

Fig. 2.

A schematic model of NF-κB activation. Various stimuli, including the proinflammatory cytokines TNF-α and IL-1, activate IKK through the action of as yet unidentified components. Once activated, IKK phosphorylates IκBα, leading to its recognition by the β-TrCP-like component of a specific ubiquitin-protein ligase that results in the polyubiquitinylation of IκBα. This then targets IκBα for rapid degradation by the 26S proteasome. IκBα degradation exposes the nuclear localization sequence on NF-κB, resulting in its translocation to the nucleus. In the nucleus, NF-κB regulates transcription of target genes, including IκBα, which functions to terminate NF-κB activity. Some of the NF-κB target genes code for inflammatory mediators, such as TNF-α and IL-1, which can lead to recruitment of additional cells to the inflammatory response.

Atypical activation pathways.

Two other activation pathways have also been reported. One is observed as a result of hypoxia or pervanadate treatment and is believed to require phosphorylation of IκBα at Tyr42 (52,53). The exact protein tyrosine kinases that participate in this pathway are not known, but certain members of the Src family have been implicated. The subsequent dissociation of tyrosine-phosphorylated IκBα from NF-κB may be mediated by interaction with phosphoinositide 3- kinase (PI 3-kinase) and not by degradation by the 26S proteasome (53). Because Tyr42 is not conserved in other IκB family members, this pathway is specific for IκBα. It would be predicted that mice whose IκBα is replaced by IκBβ (12) would fail to activate NF-κB in response to hypoxia. The second atypical activation pathway is observed in cells exposed to short-wavelength ultraviolet (UV) radiation (254 nm). UV radiation induces IκBα degradation by the 26S proteasome, but this process is not mediated by phosphorylation of Ser32 and Ser36 or Tyr42 (54,55). The mechanism of IκBα degradation in response to UV radiation is unknown. In both of these alternative pathways, NF-κB activation is considerably slower and weaker than it is in response to prototypical activators such as TNF-α, IL-1, or LPS.

IκB-Kinase (IKK): The Key to NF-κB Activation

Identification and composition.

Once it became firmly established that the critical step in NF-κB activation was IκB phosphorylation at Ser32 and Ser36 or their equivalents, effort turned toward the identification of the specific protein kinase or kinases responsible. A cytokine-responsive protein kinase activity specific for the NH2-terminal regulatory serines of IκBα and IκBβ was identified (56,57) and may correspond to a constitutive IκB kinase activity that was also described (58,59). Although numerous other enzymes were suggested to mediate IκB phosphorylation, only the kinetics of IKK activation precisely matched the kinetics of IκBα phosphorylation and degradation after treatment of cells with several NF-κB activators, including TNF-α and IL-1 (56). The majority of IKK activity eluted at an apparent molecular size of 700 to 900 kD (56, 57,60), suggesting that it was a multicomponent protein complex. However, a small amount of activity also eluted with an apparent molecular size of ~300 kD (60). By means of protein purification, microsequencing, and molecular cloning, three components of IKK were unambiguously identified. IKKα and IKKβ (also referred to as IKK1 and IKK2) are 85- and 87-kD proteins, respectively (56, 57,60-62). The DNA sequence of IKKα was previously identified with a reverse transcription polymerase chain reaction (RT-PCR)-based cloning strategy in a search for genes related to the protooncogene myc. It was identified as a putative serine/threonine kinase of unknown function called CHUK (63). IKKα was also isolated through a two-hybrid screen as a protein that interacted with the NF-κB-inducing kinase (NIK) (61). NIK is so named because, when overexpressed, it potently activates NF-κB (64). IKKα and IKKβ are very similar proteins (50% sequence identity, over 70% sequence similarity) and contain NH2-terminal protein kinase domains and leucine zipper (LZ) and helix-loop-helix (HLH) motifs (Fig. 3). IKKα and IKKβ serve as the catalytic subunits of the IKK complex.

Fig. 3.

Schematic diagram showing the known subunits of IKK and their putative functional and structural motifs. CC, coiled coil; Helix, α-helix; HLH, helix-loop-helix; LZ, leucine zipper; ZF, zinc finger.

The third component of IKK is a 48-kD regulatory subunit named IKKγ (65), NEMO (66), IKKAP1 (67), or FIP-3 (68). IKKγ was identified through the biochemical purification of the entire IKK complex (65,67) and by genetic complementation of a cell line unresponsive to NF-κB-activating stimuli (66). Secondary structure prediction algorithms indicate that IKKγ is predominantly helical with large stretches of coiled-coil structure, including a LZ motif near the COOH-terminus, whereas the first 50 and last 50 residues of IKKγ have no obvious secondary structure (65) (Fig. 3). However, the 23 COOH-terminal residues of IKKγ have a 70% sequence identity to the COOH-terminus of FIP-2 (including the three cysteines and the histidine), which encodes a putative zinc finger motif (68). FIP proteins interact with the adenovirus protein Ad E3-14.7K, a protein that inhibits TNF-α-induced cytolysis (69). IKKγ also interacts with Ad E3-14.7K and has 27% identity to FIP-2 (68). FIP-2, however, is probably not involved in NF-κB signaling (70).

Other proteins were suggested to be components of 700- to 900-kD IKK complex, including various members of the IκB and Rel families (71, 72), MAPK/ERK kinase kinase-1 (MEKK1) (57), and NIK (61, 71). However, in each of these cases, either these proteins were not detected when other purification schemes were used or immunoblots of column profiles have showed that they do not coelute with the IKKα:IKKβ:IKKγ complex. Consequently, the question of whether these proteins are integral components of IKK remains unanswered. A 150-kD regulatory subunit named IKK-complex associated protein (IKAP) was reported to associate, along with NIK, with IKK (71). However, in other studies, association of IKAP with IKK was not confirmed (73), and IKAP appears to be the homolog of a yeast protein that functions as a transcriptional elongation factor (74,75).

Much of the difficulty in determining which proteins are legitimate IKK subunits is that all but one (65) of the published biochemical purifications of IKK have been partial and therefore have yielded numerous proteins, many of which may not be directly associated with IKK. Immunoaffinity purification of IKK on a column of immobilized antibody to IKKα resulted in the purification of a core complex consisting only of IKKα, IKKβ, and IKKγ (65). The immunoaffinity column was washed with a very high stringency buffer containing 3 M urea, and, therefore, it is probable that other less tightly associated IKK components were lost.


The stoichiometry of the known components of IKK has not yet been determined. Thus far, the majority of the 700- to 900-kD complex appears to contain roughly equivalent amounts of IKKα and IKKβ (56,65) and an indeterminate amount of IKKγ. Chemical cross-linking experiments suggest that IKKγ may exist in either a dimeric or trimeric form (65). The 300-kD complex appears to contain only IKKα and IKKβ, as recombinant IKKα and IKKβ expressed in insect cells and purified to homogeneity elute with a similar apparent size (76). Chemical cross-linking experiments indicate that the 300-kD complex corresponds to dimers of IKKα and IKKβ (76). IKKγ has only been found in the 700- to 900-kD complex (65,66). In IKKγ-deficient cell lines, complexes of IKKα and IKKβ elute on a gel filtration column with an apparent size of 300 kD (66). It is possible that IKKγ through its interactions with IKKα and IKKβ and its self-association is sufficient to account for the formation of the 700- to 900-kD complex. Alternatively, IKKγ may recruit other components. The observation that a dimer containing IKKα and IKKβ elutes with an apparent size of 300 kD makes it reasonable to assume that a dimer of dimers brought together by IKKγ might elute at 700 to 900 kD, raising the possibility that the predominant form of IKK may contain four catalytic subunits.


Although IKK has generally been considered to exist as a single complex in mammalian cells, IKK might actually correspond to a family of complexes. Studies of recombinant IKKα and IKKβ expressed in insect cells indicate that IKKα and IKKβ form both homodimers and heterodimers (67, 76,77). However, the most prevalent form of active IKK thus far observed in mammalian cells contains both catalytic subunits (56, 57, 65,67). Inspection of the LZ sequences of IKKα and IKKβ fails to reveal any reason for preferential heterodimer formation (60) and indeed in IKKα- and IKKβ-deficient cell lines, homodimeric forms of the remaining catalytic subunit are readily detected complexed with IKKγ (78-82). A complex of large molecular size that appears to contain only IKKβ without IKK( was detected in HeLa cells but not in the HTLV-1 in vitro-transformed T cell line SLB (67). This complex, however, was not activatable by TNF or IL-1 (67). It is somewhat surprising that the IKKβ-only form was not detected in the earlier biochemical purification of IKK from HeLa cells (56,65). It is possible that these differences are a result of different extraction procedures. The complexes containing IKKβ only may be compartmentalized differently from the IKKα-IKKβ complexes or more simply be a pool of partially assembled IKK.

It is also possible that even the apparently homogeneous 700- to 900-kD IKK complexes are in fact heterogeneous in their content or arrangement of the two catalytic subunits. Further heterogeneity may result from association with other protein subunits. There may exist a number of different regulatory subunits that generate classes of IKK that respond to specific upstream stimuli. This would also explain some of the difficulty in unambiguously identifying other IKK components if they were present in substoichiometric amounts with respect to IKKα, IKKβ, and IKKγ.

Complex assembly.

There are several unique features of IKK that distinguish it from other known kinases such as the presence of HLH and LZ motifs. Mutational analysis indicates that the interaction between the catalytic subunits is mediated by the LZ regions (60, 62,76). The dimerization is direct and does not require other proteins (76). Furthermore, dimerization is essential for kinase activity (60,76).

The regions of IKKγ that are required for complex assembly are not yet known, and some of the existing data seem contradictory. In vitro translation experiments have shown that IKKγ can interact with IKKβ but not with IKKα (65-67). However, IKK complexes in IKKβ-deficient cells contain IKKα associated with IKKγ (79, 80,82). This could indicate that an additional component mediates the interaction between IKKα and IKKγ. However, all of the in vitro association experiments were done with proteins tagged with an epitope at the NH2- or COOH-terminus. The presence of these additional amino acid sequences could interfere with association. Studies with NH2-terminal deletion mutants of IKKγ indicate that the first 235 residues contain the site of interaction with IKKβ (67). However, a deletion mutant lacking the first 134 residues of IKKγ interacts strongly with IKKβ when expressed in mammalian cells (65), thus suggesting that a critical site of interaction exists between residues 135 and 235. Consistent with this is the observation that COOH-terminal deletion mutants of IKKγ lacking the final 119 (65) or 154 (67) amino acids do interact with IKKβ. However, an IKKγ deletion mutant lacking the first 100 residues was reported not to interact with IKKβ in mammalian cells (83). This discrepancy highlights one of the risks of using deletion mutants to determine interaction sites on proteins. Randomly generated, such deletions generally do not occur between protein structural domains. Therefore, in the region of the deletion, nonnative conformations are likely to exist that will, in some cases, propagate to other regions of the protein and result in misleading results.

Kinase domain.

The kinase domains of IKKα and IKKβ are similar to those of other known serine-threonine kinases (84). The adenosine triphosphate (ATP)-binding site is highly conserved. The conserved lysine in this region (Lys44) was mutated to generate catalytically inactive forms of IKKα and IKKβ (57, 60, 62,76). In vitro kinase assays with purified recombinant IKKα and IKKβ indicated that homodimers containing two defective subunits were catalytically inactive, but heterodimers containing a single functional kinase domain still had in vitro activity (76). In mammalian cells immunoprecipitation of catalytically inactive forms of IKKα and IKKβ yields considerable TNF-α-induced activity in in vitro kinase assays (60). This activity is due to the ability of the ectopically expressed mutant form to heterodimerize with endogenous wild-type IKKα or IKKβ. However, overexpression of the catalytically inactive forms can inhibit TNF-inducible nuclear translocation of p65 (57,60) or activation of an NF-κB-dependent reporter gene (57,62). These results suggest that IKK is rate limiting for NF-κB activation, because a small reduction in kinase activity can considerably decrease the extent of IκB degradation and NF-κB activation. Indeed, Ikkβ+/- cells had only half the IKK activity of wild-type cells, and NF-κB activation was only 20 to 30% of that in wild-type cells (80).

Kinetic analyses have been done with partially purified IKK obtained from mammalian cells (57,85) or highly purified recombinant IKKα-IKKβ dimers produced in insect cells (67, 76, 77,86). Homodimers of recombinant IKKα and IKKβ efficiently phosphorylated, with the correct specificity, various IκB substrates. The kinetic constants (Km and kcat) determined vary from one report to another such that no consensus can be drawn as to the precise values. This is primarily due to differences in experimental conditions, particularly the choice of substrate and the manner in which the enzymes were prepared. However, several important conclusions can still be made from these data. One is that the COOH-terminal region of IκBα lowers the apparent Km by an order of magnitude. The phosphorylation by partially purified IKK from mammalian cells of an NH2-terminal peptide was enhanced in the presence of a peptide corresponding to residues 279 to 303 of IκBα (85). Recombinant IKKβ phosphorylated full-length IκBα more efficiently than it did an NH2-terminal peptide (67). Similar results were obtained when comparing phosphorylation of these substrates by recombinant IKKα (76). The maximum rate (Vmax) was increased when full-length IκBα complexed with p65-p50 was used as substrate (76). This preference of IKK for IκBs complexed with NF-κB over free IκBs probably contributes to the ability of newly synthesized IκB to accumulate in cells where IKK is still active (76). It also seems from these experiments that IKK can phosphorylate all IκBs that are regulated through NH2-terminal phosphorylation and that the kinetic differences in the phosphorylation of different IκB substrates are consistent with the differences in their degradation kinetics measured in intact cells.

IKK activation.

Considerable evidence indicates that IKK activation depends on phosphorylation. Purified IKK is inactivated upon incubation with Protein Phosphatase 2A (PP2A), whereas treatment of HeLa cells with the PP2A inhibitor, okadaic acid, results in activation of IKK (56). Like other protein kinases, IKKα and IKKβ contain an activation loop within their kinase domains. This region contains specific sites whose phosphorylation causes a conformational change that results in kinase activation (87). Replacement of serines 177 and 181 in the activation loop of IKKβ with alanines prevents activation of IKK (57,88). Both of these serines are phosphorylated in cells stimulated with TNF-α or IL-1 (88), and their replacement with glutamic acid (which mimics phosphoserine) results in constitutive IKK activity (57).

Although the activation loop of IKKα is identical in sequence to the activation loop of IKKβ, its phosphorylation is not critical for IKK activation by proinflammatory stimuli. Mutation of Ser176 to alanine inactivates IKKα, whereas substitution with glutamic acid leads to full activation (89). In these experiments, however, IKKα was overexpressed and was not shown to be cytokine responsive. Under conditions in which serine to glutamic acid substitutions resulted in IKKβ activation, the same substitutions in IKKα resulted in only a minimal increase in IKK activity (57). However, with exogenous IKKα and IKKβ expressed in low amounts and well incorporated into endogenous IKK complexes, conversion of Ser176 and Ser180 in IKKα to alanine, either individually or together, had no effect on TNF-α- or IL-1-induced IKK activity (88). These results are entirely consistent with the genetic analysis of IKKα and IKKβ function. The differences between these sets of experimental data may reflect the different amounts of IKKs expressed under each transfection protocol (88). Whereas phosphorylation of the IKKα activation loop is undoubtedly important for activation of that subunit, it is not essential for activation of the IKK complex by proinflammatory stimuli. The latter, however, is entirely dependent on phosphorylation of IKKβ (88). This conclusion is consistent with the results of gene-targeting experiments in mice, which indicate that IKKβ is responsible for the IκB kinase activity generated in response to exposure to proinflammatory stimuli (78-82,90).

Another region that has a critical role in IKK activation is the HLH motif. Mutations within this motif decreased IKK activity when assayed in mammalian cells by transient transfection (60) or in assays of recombinant proteins (76). The HLH mutants still dimerize through their LZ motif (60,76) and bind to IKKγ (91). Deletion of a COOH-terminal fragment containing the HLH motif also results in inactivation of IKK, but coexpression of this fragment (amino acids 558 to 756) with the truncated kinase and LZ domains (amino acids 1 to 559) of IKKβ restores kinase activity (88). The same mutations with the HLH motif that interfere with activity of the intact kinase prevent the complementation activity of the COOH-terminal fragment (88). The HLH region of IKKβ may serve as an endogenous activator of IKK in a manner similar to the function of the cyclin subunits of cyclin-dependent kinases (CDKs) (88) (Fig. 4). The activation by the HLH region, which physically interacts with the kinase domain, requires the phosphorylation of the latter.

Fig. 4.

Schematic representation of proposed model of IKK regulation by phosphorylation. Two catalytic subunits of IKK (IKKα and IKKβ) dimerize through their leucine zippers (LZ). The COOH-terminal and helix-loop-helix (HLH) regions in IKK interact with the kinase domain (KD). Phosphorylation of the two serines in the activation loop of IKKβ results its activation. Once activated, in addition to phosphorylating IκBs, IKK autophosphorylates its COOH-terminal region. When nine or more serines in the COOH-terminal region are phosphorylated, the interaction between the COOH-terminal region, which contains the HLH motif that functions as an intrinsic activator of the kinase, and the kinase domain is altered and the activity of IKK decreases. In this state, IKK is more susceptible to further inhibition by phosphatase action.

When overexpressed in mammalian cells, a number of different protein kinases activate IKK in their wild-type forms or inhibit IKK activation in their catalytically inactive forms. These include PKCζ (92) and the MAPK kinase kinase (MAPKKK) family members NIK (62, 89,93-96, MEKK1 (94-97), MEKK2 (98), MEKK3 (98), COT/TPL-2 (99), and TAK1 (100,101).

NIK was identified by means of its association with TNF receptor-associated factor 2 (TRAF2) (64) and potently activates NF-κB when overexpressed (64,89). Expression of catalytically inactive forms of NIK block activation of NF-κB (and that of IKK whenever tested) in response to virtually all experimentally investigated inducers, including TNF-α. Consequently, NIK is thought to be directly involved in the TNF-α-induced activation of NF-κB and suggested to mediate activation of NF-κB in response to other stimuli. However, NIK also interacts with other TRAF proteins, including TRAF3, which appears not to participate in NF-κB activation (102). Furthermore, the region of TRAF2 to which NIK binds can be replaced with heterologous oligomerization domains, and resulting chimeric proteins, which no longer bind NIK, can still activate IKK (103). Yeast two-hybrid and protein interaction studies indicate that NIK strongly and preferentially interacts with IKKα (62,89), and NIK may activate IKK through direct phosphorylation of the IKKα subunit (89). However, results from IKKα knockout studies demonstrated that IKKα is not required for IKK activation by TNF-α (78, 81,90). No convincing data exist showing direct interaction between endogenous IKK complexes and NIK under physiological conditions.

An understanding of NIK's physiological role came from analysis of the alymphoplasia (aly) mutation of mice. Affected mice fail to develop lymphoid organs, and when the mutant allele was mapped and cloned it was identified as a point mutation in the NIK locus (104). The aly/aly mice exhibit a phenotype similar to that of mice deficient in lymphotoxin-α (LTα, also called TNF-β) (105) or of mice that lack the lymphotoxin β receptor (LTβR) (104). LTβR is a member of the TNF receptor family (106). The aly mutation occurs in the TRAF-interacting region and does not affect NIK's catalytic activity or its ability to activate NF-κB when overexpressed (104). The aly mutation, which presumably disrupts interactions between TRAF proteins and NIK, does not affect TNF-α-induced activation of NF-κB but does block LTβR-mediated activation of NF-κB (104). Consistent with these results, analysis of NIK-deficient cells derived from Nik-/- mice has indicated that NIK does not participate in IKK activation in response to either TNF-α or IL-1 (107). It remains to be determined whether NIK participates in IKK activation in response to LTα, as suggested by analysis of the aly mutant of NIK (104).

When attempting to identify physiologically relevant IKK kinases, it must be realized that because the activation loop motif in IKKα and IKKβ is similar to those found in many MAPK kinases (MAPKK; Ser-X-X-X-Ser, where X is any amino acid) (57, 88, 89), the ability of a MAPKKK to activate IKK when overexpressed may not mean that it is a physiological activator of NF-κB. Although catalytic activity of MEKK1 is stimulated in cells treated with TNF-α and its interaction with TRAF2 correlates with the ability of TRAF2 to activate NF-κB in transient transfection experiments (103), it remains to be seen whether MEKK1-deficient mice display defects in activation of IKK or NF-κB in response to TNF-α, IL-1, or other stimuli. The ability of catalytically inactive forms of various MAPKKKs to function as dominant negatives and interfere with IKK activation is also not sufficient to prove that the kinase in question normally functions in a particular activation pathway. Unfortunately, it appears that catalytically inactive MAPKKKs can bind to IKK upon overexpression and function as potent general inhibitors of its activation. This suggests that a more effective method of providing pathway specific information about MAPKKKs would be to generate mutant MAPKKKs that, in addition to being defective in catalytic activity, are unable to bind IKK but still interact with upstream activators. However, it is likely that progress in assigning particular kinases to specific IKK- or NF-κB-activating pathways will depend on the use of genetic approaches.

Given the multiple and diverse upstream stimuli that activate NF-κB through IKK, it seems safe to predict that several physiologically relevant IKK kinases exist. So far, it appears that IKK is the most likely point of convergence for many NF-κB signaling pathways. For example, IKKγ is required for activation of NF-κB in response to TNF-α, IL-1, LPS, dsRNA, or Tax as determined by studies in IKKγ-deficient cells (66). The COOH-terminus of IKKγ is required for activation by TNF-α or IL-1 but not for basal activity (65,67), suggesting that the COOH-terminal region may be required for recognition by activators of IKK. Consistent with this hypothesis is the observation that overexpression in mammalian cells of full-length IKKγ (68,91) or deletion mutants containing the COOH-terminal region (67,91) inhibits activation of IKK and NF-κB in response to proinflammatory stimuli. However, the COOH-terminal region is not required for activation of IKK by Tax (83). Because IKKγ is thought to recruit various upstream activators (for example, IKK kinases) to the IKK complex, such results provide indirect evidence for the existence of multiple IKK activators.

The phosphorylation of the activation loop can be also mediated by IKK itself. IKKβ and IKKα exhibit strong activity in the absence of stimulation when overexpressed in insect (67, 76, 77,86) or mammalian cells (60). Whereas it is possible that some other kinase is responsible for their activation, the simplest explanation is that IKK autophosphorylates and autoactivates. Autophosphorylation at the activation loop may amplify the extent of IKK activation through an upstream kinase. Autoactivation also provides a mechanism whereby the activation of a single catalytic subunit results in the activation of all catalytic subunits within an individual IKK complex. Autophosphorylation may play a more central role in the response to certain activators, such as the Tax protein of HTLV-1 (26). Although Tax is not a protein kinase, it can function as a dimerizer that can stabilize dimer formation in other proteins (26). Consequently Tax, which has been shown to directly associate with IKK through IKKγ (83, 108,109), may bring two IKK catalytic domains into close proximity and thereby induce their autoactivation.

Inactivation of IKK.

The mechanism of IKK inactivation is not fully understood, but it is clear that most stimuli cause only transient IKK activation; the stronger the initial activation the more transient the response seems to be. Transient IKK activation and mechanisms to limit its activity are physiologically important because persistent NF-κB activity can result in deleterious or fatal conditions, such as septic shock or acute inflammation. In cells stimulated with TNF-α, IKK activity peaks within 5 to 15 min, and after ~30 min it decreases to about 25% of its peak value: Activity then decreases only slightly over the next 90 min (56, 60,88). Consistent with these kinetics, degradation of IκBα is generally complete within the first 15 min after TNF-α treatment, and resynthesized pools of cytoplasmic IκBα reappear within 60 min. Small decreases in IKK activity result in large decreases in IκB degradation and NF-κB activation. Consequently, the ability of newly synthesized IκBα to escape degradation probably results from a combination of the apparent decrease in IKK activity and the lower catalytic activity of IKK toward free IκB (76).

The initial down-regulation of IKK in TNF-α-stimulated cells appears to be regulated by COOH-terminal autophosphorylation of IKK. After treatment of cells with TNF-α, IKKα and IKKβ are heavily autophosphorylated at their COOH-terminal regions (88). Replacement of 10 autophosphorylated serines by alanines in the COOH-terminal portion of IKKβ prolongs IKK activation to nearly four times that of the wild-type enzyme. Conversely, replacement of these serines by glutamic acid, which mimics phosphoserine, resulted in a greatly diminished amount of TNF-α-induced activity of IKK (88). Because inhibition requires multiple phosphorylation events that may occur more readily after IκBs, which compete for IKK phosphorylation, have been depleted, the COOH-terminal autophosphorylation can serve as a timing mechanism to limit the duration of strong IKK activation. Phosphorylation of many of the 10 residues may result in a conformational change of the COOH-terminal region, which contains the HLH motif, leading to a decrease in kinase activity (88) (Fig. 4). An alternative possibility is that IKK is down-regulated by a phosphatase that is recruited to the phosphorylated COOH-terminal region. A conformational change could decrease the efficiency of activation loop autophosphorylation, thus rendering the kinase more sensitive to phosphatase action. A substantial role for constitutive phosphatases in inactivation of IKK is suggested by the ability of the phosphatase inhibitor okadaic acid to activate IKK in mammalian cells (56).

Physiological function of IKK.

Given the high sequence similarity between IKKα and IKKβ and the fact that recombinant forms of both kinases effectively phosphorylate IκB in vitro (67, 76, 77,86), it seemed likely that the two kinases might have functionally redundant roles in NF-κB activation. The presence of both of these protein kinases in a single complex made it even more difficult to determine whether they had distinct functions and, if so, to identify those functions. This issue seems less complex for IKKγ. It appears to be the only stoichiometric regulatory subunit of the IKK complex, and IKKγ-deficient cells are fully defective in activation of IKK and NF-κB in response to numerous stimuli, including TNF-α, IL-1, dsRNA, LPS, and Tax (66).

Analysis of phosphorylation of IKKα and IKKβ suggested that only IKKβ has a critical role in the response to proinflammatory stimuli and that IKKα may serve a different function (88). However, the functional distinction between the two catalytic subunits became clear only after the generation of IKKα- and IKKβ-deficient mice through gene-targeting experiments. The phenotype of the IKKα-deficient mice was unexpected. Ikkα -/- mice are born alive but die within 4 hours after birth (78,81,90). Superficially, newborn Ikkα-/- mice appear to lack limbs, tails, and ears. They also exhibit severe cranial and facial deformities, and their skin is smooth, taut, and very shiny. Further analysis revealed that Ikkα-/- mice contain most of the skeletal elements of the limbs and tail, as well as apparently normal limb musculature. However, their limbs and tail are glued onto the rest of the body and covered by the thickened skin. Although the proximal limb bones were almost normal in size, ossification, and morphology, some of the distal elements (digits) were missing, and others were fused. This aberration is probably caused by an earlier defect in interdigital apoptosis, which results in formation of webbed limb buds (78). The most dramatic cellular defect in Ikkα-/- mice is in formation and organization of the epidermis. In the mutants, the epidermis is up to 5 to 10 times as thick as that of normal mice, a defect that is due to excessive proliferation of the basal layer. Similar thickening of the suprabasal layer was observed in transgenic mice that express dominant negative IκB proteins (110). In addition, there is almost a complete absence of epidermal differentiation; instead of having well stratified epithelium composed of four cell types, the epidermis of Ikkα-/- mice is more or less uniform. Because of the absence of the two upper layers, the stratum corneum and stratum granulosum, the mutant epidermis is very sticky, thereby explaining why the limbs are fused to the body. Thus, IKKα has a major role in epidermal differentiation.

IKKα was revealed not to be required for IKK activation in response to proinflammatory stimuli. Stimulation of Ikkα-/- embryonic fibroblasts, primary keratinocytes, or liver tissue with TNF-α, IL-1, or LPS resulted in normal IKK activation and IκBα degradation (78,81). NF-κB DNA-binding activity is however reduced by ~50% in Ikkα-/- embryonic fibroblasts, suggesting that IKKα has some as yet undetermined role in NF-κB activation (78,81).

Although these results suggest that IKKα is not critical for activation of NF-κB by proinflammatory stimuli, it appears that IKKα is critical for IκB-dependent activation of NF-κB in response to an as yet unidentified developmental signal that triggers keratinocyte differentiation. NF-κB appears to be activated during keratinocyte differentiation, such that p65/RelA is cytoplasmic in basal cells but nuclear in their differentiated derivatives (110,111). This process could be coupled to IκBα degradation, because the abundance of IκBα is lower in differentiated keratinocytes than in basal cells. In Ikkα-/- mice, however, IκBα is highly abundant throughout the epidermis, and p65 remains cytoplasmic (111). The signals that control this process are unknown, but this developmental control of NF-κB activity resembles the control of the transcription factor Dorsal in Drosophila (112). One of the target genes for Dorsal is twist (113,114). Defects in the mammalian Twist gene lead to craniofacial, limb, and skeletal abnormalities in humans (115,116). Reduced Twist expression was also observed in Ikkα-/- fetuses (90). However, it is not clear to what extent the decreased expression of Twist contributes to any of the morphogenetic defects in Ikkα-/- mice. Despite the defects in activation of NF-κB during development, Ikkα-/- keratinocytes are highly responsive to IL-1 and TNF-α and efficiently activate NF-κB in response to these stimuli (111). This response is most likely mediated by IKKβ.

The IKKβ-deficient embryos die at embryonic day (E) 12.5 to 14.5 as a result of massive apoptosis in the liver (79, 80,82). An essentially identical phenotype is presented by mice deficient in p65/RelA, which die at E14.5 (117), or mice that are deficient in both p65 and p50, which die at E12.5 (118). Expression of p65 is required to protect liver cells from TNF-α-induced apoptosis, and mice that lack both TNF-α and p65 are viable and have normal livers (119). For unknown reasons, embryonic liver expresses copious amounts of TNF-α, which in the absence of NF-κB activity triggers massive apoptosis. Consistent with this hypothesis is the observation that IKKβ-deficient cells are more sensitive to TNF-α-induced apoptosis than are wild-type cells (79,82). Ikkα-/- cells exhibit very little activation of IKK or NF-κB in response to TNF-α or IL-1 (79, 80,82). Furthermore, even Ikkβ+/- cells, which express half as much IKKβ as do normal cells, exhibit a 50% decrease in kinase activity of IKK and a 70 to 80% decrease in NF-κB activation (80). Undoubtedly, IKKβ is the catalytic subunit required for NF-κB activation in response to proinflammatory stimuli.

Collectively, the results of the IKKα and IKKβ knockout studies demonstrate that despite the predominant form of IKK being a complex that contains both IKKα and IKKβ, homodimeric complexes are able to function effectively. Ikkα-/- mice have apparently normal livers and spleens, whereas Ikkβ-/- mice, although defective in response to proinflammatory stimuli, exhibit none of the skin or skeletal abnormalities of the Ikkα-/- mice. Thus far, Ikkα-/- and Ikkβ-/- mice appear to have no common phenotype that could provide a functional explanation for formation of complexes containing both IKKα and IKKβ.

Signaling to IKKs by Proinflammatory Cytokines

The proinflammatory cytokines TNF-α and IL-1 are important mediators of inflammation and potent activators of NF-κB. They exert their effects through interaction with cell surface receptors. TNF-α has two distinct receptors, TNFR1 and TNFR2, but in most cell types NF-κB activation occurs primarily through TNFR1 (106, 120,121). Similarly, of the two known IL-1 receptors, only IL-1R1 participates in cellular signaling (122,123). Although the biological functions of TNF-α and IL-1 are remarkably similar, the two cytokines and their receptors are not related in sequence. Binding of ligand to either receptor induces receptor clustering and results in recruitment of signaling proteins to the receptor cytoplasmic domains. TNFR1 interacts with TRADD (TNF receptor 1-associated death domain protein) which functions as an adapter for the recruitment of other proteins including RIP (receptor-interacting protein), a serine-threonine kinase, and TRAF2 (106, 120,121). Studies of RIP-deficient cells indicate that RIP is required for NF-κB activation in response to TNF-α but not activation in response to IL-1 or LPS (124). Activation of NF-κB through IL-1R1 requires a membrane-spanning accessory protein, IL-1RACP (125,126). This complex along with the adaptor protein MyD88 recruits a serine-threonine kinase, IRAK (IL-1R-activated kinase) (126-128). IRAK then participates in the recruitment of a protein related to TRAF2 called TRAF6 (128,129). Thus, the common feature in signaling through TNFR1 and IL-1R is the role of TRAF proteins.

TRAF2 and TRAF6, like other members of the TRAF family (121, 130,131), contain a common region, the TRAF domain, in their COOH-terminal half. The TRAF domain can be further divided into two subdomains: TRAF-C, which is highly conserved and mediates heterodimerization and homodimerization of TRAF proteins as well as interactions with receptors, and TRAF-N, which is more variable and has a coiled-coil structure (132). Most TRAF proteins contain a NH2-terminal RING finger and several zinc-finger motifs. Overexpression of either TRAF2 or TRAF6 results in activation of NF-κB, and it appears that the NH2-terminal region is required for this activity (102, 129,133-135). The NH2-terminal portions of TRAF2 and TRAF6 are sufficient for activation of downstream effectors when fused to a foreign protein domain that can be inducibly oligomerized (103). Oligomerization of these chimeras results in efficient activation of IKK and NF-κB, which depends on the integrity of the NH2-terminal region, especially the RING finger (103).

The unique role of individual TRAF family members in specific pathways is not thoroughly defined. Cells isolated from TRAF2 knockout and transgenic mice indicate that TRAF2 is not required for activation of NF-κB in response to TNF-α (136,137). One possible explanation for this observation is that other TRAF proteins, such as TRAF5 (138), may substitute for TRAF2. In cells obtained from TRAF6 knockout mice, NF-κB activation was severely impaired in response to IL-1 or LPS treatment (139). However, no defects in NF-κB activation in response to TNF-α treatment were observed in these cells. The requirement of TRAF6 for LPS-induced activation of NF-κB suggests that TRAF6 is deployed by members of the Toll receptor family (139). The ability of individual TRAF proteins to mediate NF-κB activation through interaction with various receptors is further demonstrated by studies of Traf6-/- mice, which indicate that TRAF6 is required for signaling through CD40 and receptor activator of NF-κB (RANK), two members of the TNFR family (139).

The mechanism by which TRAF activates IKK is not understood, although presumably clustering of the NH2-terminal domain forms a recognition site for a downstream signaling protein. MEKK1 interacts with the effector domain of TRAF2 in a manner that depends on oligomerization of TRAF2, but this event was suggested to be important for activation of MAP kinase cascades rather than activation of IKK (103). Another potential target for TRAF proteins is NIK (64,102). However, NIK does not bind to the NH2-terminal effector domain (103) and as discussed above is not required for activation of IKK in response to TNF-α or IL-1. Other potential targets include germinal center kinase (GCK) (140), GCK-related kinase (GCKR) (141), and apoptosis signal regulating kinase-1 (ASK1) (142,143). However, ASK1 and GCK appear not to interact with the NH2-terminal effector domain (103). All of these kinases are members of the MAPKKK group. Given that many MAPKKKs can activate IKK, members of this group could be the missing pieces in the cytokine-inducible NF-κB signaling pathways, connecting TRAF proteins to IKK.

Ubiquitinylation and Degradation of IκB

IKK-mediated phosphorylation of the NH2-terminal regulatory region of IκB induces its rapid polyubiquitinylation. The principal sites of ubiquitinylation in IκBα are lysines 21 and 22 and their substitution by arginine impairs degradation (39,45-47). Protein ubiquitinylation is a three-step process (144). First, ubiquitin is activated through an ATP-dependent process by E1, the ubiquitin-activating enzyme. Second, the activated ubiquitin is transferred to cysteine in the active site of an E2 enzyme, the ubiquitin-carrier protein. The third step, the transfer of the activated ubiquitin to the target protein, is catalyzed by an E3 enzyme, the ubiquitin ligase. Generally, there is only a single form of E1, many forms of E2, and a very large number of E3s. E3s exist in many separate families. The E3 enzyme provides specificity in recognition of the substrate. In a cell-free system, ubiquitinylating activity specific for NH2 terminally phosphorylated IκB is constitutive (145). The site of E3 recognition is functionally conserved between IκBα and IκBβ and does not involve the nearby ubiquitinylation site (145). The E3 component required for recognition of IκBα was identified and found to correspond to β-TrCP or β-TrCP variants (42, 146,147). β-TrCP is a member of the F-box/WD-repeat family and exists as part of a complex with S phase kinase-associated protein 1 (Skp1) and a member of the Cullin family (148,149). β-TrCP recognizes sequences very similar to those at the phosphoacceptor sites of IκBs, including similar sequences in β-catenin and the HIV protein, Vpu (43). This observation coupled with results of peptide competition studies (42,145) indicates that the recognition of IκB by β-TrCP is not due to a phosphorylation-induced global conformational change in IκB but rather that the highly conserved region around the IκB phosphoacceptor sites serves as a recognition site for both IKK and β-TrCP.

Although the ubiquitinylation machinery is constitutively active, it does offer the possibility that specific components may be inhibited by regulatable factors. As yet, however, such regulation has not been described. Another way in which ubiquitinylation can be disrupted is by blocking the site of ubiquitinylation or ubiquitin chain elongation. One such ubiquitinylation antagonist, SUMO-1, a small ubiquitin-like protein, was found linked to Lys-21 of IκBα (150). Modification by SUMO-1 blocks ubiquitinylation at that site and results in an IκBα that is resistant to signal-induced degradation (150). Unfortunately, the manner by which SUMO-1 modification of IκBα is regulated is unknown, and, therefore, its physiological importance remains to be determined.

Both IκBα and IκBβ contain a COOH-terminal region enriched in proline (P), glutamic acid (E), serine (S), and threonine (T). These PEST sequences are frequently found in proteins that undergo rapid degradation (151). The role of the PEST region in signal-induced degradation of IκBα is controversial. A number of studies suggest that the PEST region contributes to induced degradation of IκBα (37, 41, 152,153), but others have reached the opposite conclusion (154-156). A simple explanation that reconciles these contradictory results is that the rate of signal-induced degradation of various mutant and chimeric forms of IκBα reflects the rate at which their NH2-terminal region is phosphorylated by IKK. Forms of IκBα that bind more weakly to NF-κB are degraded more slowly (154,156), which is consistent with the preference of IKK for IκBα complexed with NF-κB over free IκBα (76). There is also a second factor that affects NH2-terminal phosphorylation of IκBα in these studies: The presence of the PEST region enhances IκB kinase activity (85). Thus, it appears that after the signal-induced NH2-terminal phosphorylation of IκBα, the PEST region does not contribute to the rapid degradation of IκBα by the proteasome.

Processing of p100/p105: An Additional Point of Control?

The precursor proteins, p105 and p100 (or NF-κB1 and NF-κB2), undergo limited proteolytic processing to yield p50 and p52, respectively. The precise COOH-termini of p50 and p52 have not been determined by protein sequencing, but deletion studies suggest that p50 contains the first 433 amino acids of p105 (157) and that p52 contains the first 405 amino acids of p100 (158). Like degradation of IκB, this process relies on ubiquitinylation and is mediated by the 26S proteasome (159,160). However, processing of p105 and p100 is relatively slow, and it represents an unusual example of limited proteolysis by the proteasome (161,162). Unprocessed, full-length p100 and p105 form stable dimers with other Rel proteins and cause their cytoplasmic retention through their COOH-terminal ankyrin domains (163-165). Consequently, partial or complete degradation of p100 or p105 results in nuclear translocation of NF-κB. Increased processing of p100 and p105 occurs in response to NF-κB-activating stimuli (164,166-168), but the mechanism of such regulation has not been determined. The extent by which processing of p105 and p100 is accelerated in response to proinflammatory cytokines is not very great, and it is not clear to what extent it contributes to regulation of NF-κB activity. Most studies on processing of NF-κB precursors have been done with p105, but it is generally assumed that processing of p100 occurs by a similar mechanism. An alternative pathway in which cotranslational processing of p105 yields p50 was proposed to be the principle source of p50 (169). However, pulse chase experiments in both COS-7 cells and in cell-free systems indicate that most p50 is produced posttranslationally from the p105 precursor (162).

A 23-amino acid glycine-rich region (GRR) of p105 appears to serve as a signal for limited proteolysis (157). Removal of the GRR prevents formation of p50 formation, whereas insertion of the GRR into proteins unrelated to NF-κB is sufficient to promote processing similar to that of wild-type p105 (157). This result suggested that the GRR might act as a recognition site for an endoproteolytic enzyme that separates the NH2-terminus, p50, from the ubiquitinylated COOH-terminus, which is then degraded (157). This process would presumably be mediated by a proteasome-associated enzyme because in vitro reconstitution has demonstrated that the 26S proteasome contains all of the proteolytic activities required for proper p105 processing (161). The GRR, however, is unlikely to serve as a universal recognition site for endoproteolytic enzymes, because recent studies indicate that only a small subset of GRR chimeras are properly processed by the proteasome (162). This has led to an alternative and more likely explanation of the GRR's function: that it acts as a physical barrier to prevent entry of p50 into the cavity of the proteasome, thereby preventing further processing (162). This explanation is supported by data from a cell-free system, in which GRR-containing p50 was more resistant to degradation than p50 lacking the GRR (162). A similar function was proposed for the GRR in the cotranslational processing of p105 by the proteasome (169).

The COOH-terminal PEST region of p105 has an important role in p105 processing. Removal of the final 69 to 89 residues results in more efficient processing (161,167). Enhancement of processing in cells treated with phorbol esters and ionomycin requires phosphorylation in that same COOH-terminal region (167), and mutation of serines in that region retards basal p105 processing (170). These results suggest that the COOH-terminal PEST region inhibits p105 processing unless it is phosphorylated (161,167). The kinases responsible for this phosphorylation are not known, although it has been suggested on the basis of similarities between the PEST regions of p105 and IκBα that Casein Kinase II (CKII), a constitutively active protein kinase, could be involved and that regulation might result from activity of a signal sensitive phosphatase (167).

Another protein kinase, TPL-2, associates with p105 and stimulates its proteasome-mediated degradation, resulting in NF-κB activation (171). TPL-2 is a member of the MAPKKK family and can activate IKK (99). Consequently, much of the nuclear translocation of NF-κB observed in cells overexpressing TPL-2 may result from degradation of IκBα. Moreover, in cells overexpressing TPL-2, most p105 is completely degraded and not processed to p50 (171), making the physiological importance of TPL-2 binding to p105 unclear. This is a common problem in many studies of p105 processing, as conclusions are often drawn on the basis of disappearance of p105 and not the appearance of p50. Consequently, limited proteolysis and complete degradation are often not distinguished. Complete degradation of p105 was reported after stimulation of monocytes by LPS (172). This further suggests that amounts of p105 may be regulated by two proteolytic pathways, limited and complete. Because both LPS and TPL-2 are activators of IKK, it is possible that IKK could promote degradation of p105. However, the time course of degradation of p105 is slow, so it is also possible that its degradation is enhanced by products of NF-κB-responsive genes.

Rel/NF-κB Activation in Drosophila

The NF-κB signaling pathway is highly conserved evolutionarily, and the role of Drosophila Rel or NF-κB proteins in development has been well characterized (112,173). However, understanding of the specific signaling events that regulate activation of the Drosophila Rel or NF-κB proteins, Dorsal, Relish, and Dif lags behind our knowledge of NF-κB activation pathways in mammalian systems. The signaling pathway for Dorsal activation is the most thoroughly studied (174,175) and is remarkably similar to the IL-1 signaling pathway. An extracellular ligand binds to Toll, whose intracellular domain is similar to that of the IL-1 receptor, and results in the recruitment of the IRAK homolog, Pelle, through the novel adapter protein Tube. This results in phosphorylation and degradation of the IκB protein, Cactus. As of yet, no Drosophila homologs of IKK have been reported. It is also not clear what the actual targets of Pelle's kinase activity might be. Indeed, kinase activity of its mammalian homolog IRAK is not required for IL-1 signaling (176-179). A family of mammalian Toll-like proteins has been identified (180). These proteins may function as signaling receptors for LPS and other bacterial endotoxins (180,181). This function is related to the second function of Drosophila Toll, which is the activation of innate immune responses (181). The signaling pathway by which Toll activates innate immune responses is essentially identical to the pathway that activates Dorsal during development.

Nuclear translocation of Dif, like that of Dorsal, depends on the degradation of cactus (182). However, activation of Dif is apparently largely independent of Pelle and Tube, because Dif translocates normally in fat-body cells from Toll-/- and pelle-/- larvae (182). Dif appears to be regulated primarily through 18-Wheeler, a member of the Toll receptor family (183). However, Dif, or more likely Dif-containing heterodimers, are activated by Toll (184). Little is known about the processing or activation of Relish, a p100/p105 homolog (185), although Relish:Dorsal and Relish:Dif complexes appear to be important for the Drosophila immune response (186).

Regulation of NF-κB Activity not Mediated by IκB Degradation

Although degradation of IκB is sufficient to cause nuclear translocation of NF-κB, subsequent events can affect NF-κB's ability to activate transcription. Although the presence and activity of other nuclear proteins, including transcription factors, such as AP-1, and components of the basal transcriptional machinery, certainly contribute to NF-κB's transcriptional activity, such considerations are outside the scope of this review. However, it appears that phosphorylation of NF-κB proteins may modulate their transcriptional activity. There are numerous reports of NF-κB being phosphorylated in response to activating stimuli (187-194). Treatment of cells with LPS induces the phosphorylation of Ser276 of p65 by protein kinase A (PKA), and this phosphorylation increases the transcriptional activity of p65 by strengthening the interaction between p65 and the transcriptional coactivator CBP/p300 in the nucleus (194).

Stimulation of cells with TNF-α results in phosphorylation of Ser529 of p65 (193). Although this phosphorylation increases p65 transcriptional activity, it has no effect on nuclear translocation or DNA binding. The kinase responsible for this phosphorylation has not been identified. IKK purified from mammalian cells phosphorylated p65 in vitro, but this activity was distinct from IκB kinase activity in that it could be dissociated from IKK by stringent washing (57). Purified recombinant IKKα/β can also phosphorylate p65 (67). It is not known if the phosphorylation of p65 by IKK is physiologically relevant.

An additional modulation of NF-κB activity is suggested by observations made on IKKα-deficient cells (78,81). Ikkα-/- embryonic fibroblasts exhibit substantially lower amounts of both basal and induced NF-κB DNA-binding activity than do wild-type cells. However, TNF-α-induced IκB degradation and in vitro IκB kinase activity are similar in wild-type and Ikkα-/- fibroblasts. The basis for the difference in DNA-binding activity is not yet known. It may result from IKKα-mediated phosphorylation of an NF-κB component or from an effect on the processing of p105 to p50.

Serine phosphorylation in the central portion of Dorsal is required for its signal-induced nuclear localization (195). Signal-induced phosphorylation occurs on Ser317 of Dorsal while still bound to Cactus in the cytoplasm. Although mutation of Ser317 to alanine in Dorsal blocks nuclear localization in response to the ventral signal, it does not appear to affect signal-induced degradation of Cactus or basal nuclear translocation of Dorsal. Ser317 is completely conserved among Rel/NF-κB members, and this phosphorylation could be a general mechanism of regulating signal-induced nuclear translocation of NF-κB (195).

Termination and Down-Regulation of NF-κB Activity

Negative regulation of NF-κB activity is very complex and several mechanisms lead to termination of NF-κB activation or its down-regulation in response to specific signals. IKK inactivation is required to prevent or delay its reactivation and possible mechanisms of such regulation were discussed above. The critical inhibitory step, however, is thought to be binding of newly synthesized IκBα to NF-κB in the nucleus. It is not known whether IκBα actively removes NF-κB from the DNA or whether additional factors contribute to the process. The NF-κB:IκBα complexes are then exported to the cytoplasm by means of an NES present on IκBα (9). Inhibition of Crm1, a member of the karyopherin family, blocks NES-mediated nuclear protein export, causing IκBα to accumulate in the nucleus (196). This results in decreased NF-κB activity and nuclear retention of NF-κB:IκBα complexes following stimulation by TNF-α (196). Nuclear localization of IκBα may also be regulated. Nuclear translocation of IκBα was originally thought to be a passive process, but potential NLS regions were identified (197) and IκBα translocation is an energy-dependent process (198). In vitro studies suggest that an as yet unidentified IκBα-specific carrier protein may be required for translocation (198).

Another aspect of NF-κB down-regulation by IκBs is the phosphorylation of residues in the COOH-terminal region of IκB. Although it appears that constitutive phosphorylation of IκBα contributes to regulation of its half-life in unstimulated cells (199-201), constitutive COOH-terminal phosphorylation of IκBβ appears to have a more important role in NF-κB regulation, specifically in controlling persistent activation of NF-κB. Prolonged exposure to certain stimuli, such as LPS, leads to long-term induction of NF-κB activity, in spite of large amounts of newly synthesized IκBα. The presence of newly synthesized unphosphorylated IκBβ in the nucleus, which in contrast to COOH terminally phosphorylated IκBβ does not inhibit the DNA-binding ability of NF-κB or mask its NLS, may protect NF-κB from inhibition by IκBα (202,203). In order for IκBβ to function in this manner, it must remain bound to NF-κB in the NF-κB-DNA complex, and experimental results indicate that it does so (203-206). Structural studies of the NF-κB:IκBα complex suggest that sequence differences in the sixth ankyrin repeats of IκBα and IκBβ might account for this functional difference between IκBα and IκBβ (17). However, the replacement of IκBα coding sequences with those of IκBβ did not result in any pathophysiological aberrations or changes in NF-κB regulation (12). Thus, there is little genetic support for any substantial difference in the physical properties of these two IκB proteins.


Considerable progress has been made in understanding the mechanisms of NF-κB activation, in particular the identification and initial characterization of the IKK and β-TrCP-like ubiquitin ligase complexes. However, considerable work is still required to obtain a complete understanding of the composition, function, assembly, and regulation of both complexes. The direct upstream activators of IKK and their mechanisms of action also offer a fertile ground for future research. It is likely that in addition to positive regulators, there are also negative regulators of IKK activity, as befitting such a critically important enzyme. New results should continue to provide fascinating surprises and generally applicable lessons. Finally, it must be reiterated that NF-κB is actually a family of proteins, as are the inhibitory molecules that dictate their subcellular localization. Therefore, considering all of the conceivable ways these proteins can combine, there are well over 100 possible forms of NF-κB:IκB complexes that could exist in cells. This tremendous diversity has been for the most part ignored, and consequently, much of the subtlety of NF-κB signaling remains to be explored. It is becoming clear, however, that specific NF-κB complexes have distinct regulatory functions, and genetic evidence from Drosophila indicates that the form of NF-κB activated may depend on the specific stimulus.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
  71. 71.
  72. 72.
  73. 73.
  74. 74.
  75. 75.
  76. 76.
  77. 77.
  78. 78.
  79. 79.
  80. 80.
  81. 81.
  82. 82.
  83. 83.
  84. 84.
  85. 85.
  86. 86.
  87. 87.
  88. 88.
  89. 89.
  90. 90.
  91. 91.
  92. 92.
  93. 93.
  94. 94.
  95. 95.
  96. 96.
  97. 97.
  98. 98.
  99. 99.
  100. 100.
  101. 101.
  102. 102.
  103. 103.
  104. 104.
  105. 105.
  106. 106.
  107. 107.
  108. 108.
  109. 109.
  110. 110.
  111. 111.
  112. 112.
  113. 113.
  114. 114.
  115. 115.
  116. 116.
  117. 117.
  118. 118.
  119. 119.
  120. 120.
  121. 121.
  122. 122.
  123. 123.
  124. 124.
  125. 125.
  126. 126.
  127. 127.
  128. 128.
  129. 129.
  130. 130.
  131. 131.
  132. 132.
  133. 133.
  134. 134.
  135. 135.
  136. 136.
  137. 137.
  138. 138.
  139. 139.
  140. 140.
  141. 141.
  142. 142.
  143. 143.
  144. 144.
  145. 145.
  146. 146.
  147. 147.
  148. 148.
  149. 149.
  150. 150.
  151. 151.
  152. 152.
  153. 153.
  154. 154.
  155. 155.
  156. 156.
  157. 157.
  158. 158.
  159. 159.
  160. 160.
  161. 161.
  162. 162.
  163. 163.
  164. 164.
  165. 165.
  166. 166.
  167. 167.
  168. 168.
  169. 169.
  170. 170.
  171. 171.
  172. 172.
  173. 173.
  174. 174.
  175. 175.
  176. 176.
  177. 177.
  178. 178.
  179. 179.
  180. 180.
  181. 181.
  182. 182.
  183. 183.
  184. 184.
  185. 185.
  186. 186.
  187. 187.
  188. 188.
  189. 189.
  190. 190.
  191. 191.
  192. 192.
  193. 193.
  194. 194.
  195. 195.
  196. 196.
  197. 197.
  198. 198.
  199. 199.
  200. 200.
  201. 201.
  202. 202.
  203. 203.
  204. 204.
  205. 205.
  206. 206.
  207. 207.
View Abstract

Navigate This Article