IKK-1 and IKK-2: Cytokine-Activated IκB Kinases Essential for NF-κB Activation

See allHide authors and affiliations

Science  31 Oct 1997:
Vol. 278, Issue 5339, pp. 860-866
DOI: 10.1126/science.278.5339.860


Activation of the transcription factor nuclear factor kappa B (NF-κB) is controlled by sequential phosphorylation, ubiquitination, and degradation of its inhibitory subunit IκB. A large multiprotein complex, the IκB kinase (IKK) signalsome, was purified from HeLa cells and found to contain a cytokine-inducible IκB kinase activity that phosphorylates IκB-α and IκB-β. Two components of the IKK signalsome, IKK-1 and IKK-2, were identified as closely related protein serine kinases containing leucine zipper and helix-loop-helix protein interaction motifs. Mutant versions of IKK-2 had pronounced effects on RelA nuclear translocation and NF-κB–dependent reporter activity, consistent with a critical role for the IKK kinases in the NF-κB signaling pathway.

Transcription factors of the NF-κB Rel family are critical regulators of genes that function in inflammation, cell proliferation, and apoptosis (1). The prototype member of the family, NF-κB, is composed of a dimer of p50 (NF-κB1) and p65 (RelA) (2). NF-κB exists in the cytoplasm of resting cells but enters the nucleus in response to various stimuli, including viral infection, ultraviolet irradiation, and proinflammatory cytokines such as tumor necrosis factor α (TNF-α) and interleukin-1 (IL-1) (1, 3).

Activation of NF-κB is controlled by an inhibitory subunit, IκB, which retains NF-κB in the cytoplasm (4). NF-κB activation requires sequential phosphorylation, ubiquitination, and degradation of IκB as well as consequent exposure of a nuclear localization signal on NF-κB (5). Ser32 and Ser36 of IκB-α, and the corresponding Ser19 and Ser23 of IκB-β, represent critical phosphorylated residues (6). The IκB kinase shows a high degree of specificity for these residues, because an IκB-α variant in which Ser32 and Ser36 were substituted by Thr (S32T, S36T) showed much reduced phosphorylation and degradation in stimulated cells and interfered with endogenous NF-κB activation (6).

To identify the IκB kinase responsible for the initial critical step of NF-κB activation, we fractionated whole-cell extracts (WCEs) from TNF-α–stimulated HeLa cells by standard chromatographic methods (7). We assayed IκB kinase activity in each fraction by phosphorylating glutathione-S-transferase (GST)–IκB-α (1–54) or GST–IκB-β (1–44) (8). Kinase specificity was established by using (S32T, S36T) mutant GST–IκB-α (1–54) [GST–IκB-α (1–54; S32T, S36T)], and GST–IκB-β (1–44), in which Ser19 and Ser23 were mutated to Ala [GST–IκB-β (1–44; S19A, S23A)] (8). IκB kinase activity was not observed in unstimulated cell extracts but was strong in cells stimulated for 5 to 7 min with TNF-α (9). Gel-filtration chromatography resolved this IκB kinase activity in a broad peak of 500 to 700 kD (Fig. 1A). In contrast to the 600-kD IκB kinase complex that was observed after treatment of cell extracts with either okadaic acid or ubiquitin-conjugating enzymes (10), the IκB kinase activity described here displayed no requirement for ubiquitination (9). We refer to the protein complex that contains the inducible IκB kinase activity as the IKK signalsome.

Figure 1

Identification of the IKK signalsome. (A) IκB-α kinase activity chromatographs as a large complex (500 to 700 kD). WCE of TNF-α–stimulated (20 ng/ml; 7 min) HeLa S3 cells was prepared, fractionated on a Superdex 200 gel-filtration column, and monitored for IκB-α kinase activity (8). Phosphorylation of the GST–IκB-α (1–54) WT substrate is indicated by arrow on the right. Molecular mass standards are indicated by arrows on top. (B) Identification of proteins that cochromatograph with the IKK signalsome. IKK signalsome was partially purified from extracts of TNF-α–stimulated HeLa S3 cells by sequential fractionation on Q Sepharose, Superdex 200 gel-filtration, Mono Q, and phenyl Superose columns. Phenyl Superose fractions containing the peak of IKK signalsome activity were subjected to Western blot analysis with several different antibodies, indicated on the left. The relative level of IKK signalsome activity is indicated by the number of plus signs in upper shaded area.

NF-κB activation occurs under conditions that also stimulate mitogen-activated protein kinase (MAP kinase) pathways (11). We tested preparations containing the IKK signalsome for the presence of proteins associated with MAP kinase and phosphatase cascades (Fig. 1B). The MAP kinase kinase–1 (MEKK-1) and two Tyr-phosphorylated proteins of ∼55 and ∼40 kD copurified with IκB kinase activity (Fig. 1B). A protein of ∼50 kD that reacted with an antibody to MAP kinase phosphatase–1 (anti–MKP-1) also copurified with the IκB kinase through several purification steps.

We examined antibodies against proteins copurifying with the IKK signalsome activity for their ability to immunoprecipitate IκB kinase activity. Of a panel of antibodies tested, one of three anti–MKP-1 efficiently coimmunoprecipitated an inducible IκB kinase activity from HeLa cells (12) and primary human umbilical vein endothelial cells (HUVECs) (9). IκB kinase activity was not detected in immunoprecipitates from unstimulated HeLa cells, but it was detected within minutes of exposure of cells to TNF-α (Fig.2A). This IκB kinase did not phosphorylate GST–IκB-α (1–54; S32T, S36T). IκB-α kinase activity was maximal by 5 min and declined thereafter, consistent with the time course of IκB-α phosphorylation and degradation (Fig. 2A). Kinase activity was also induced by stimulation of cells with IL-1 or phorbol 12-myristate 13-acetate (PMA) (Fig. 2B); moreover, no increase in activity was detected from HeLa cells treated with N-tosyl-l-phenylalanine chloromethyl ketone (TPCK), an inhibitor of NF-κB activation (6), before cell stimulation with TNF-α.

Figure 2

Biochemical properties of IKK signalsome activity immunoprecipitated by anti–MKP-1. (A) Increased activity of IKK signalsome after cell stimulation. (Upper) Time course for increased IKK signalsome activity. Anti–MKP-1 immunoprecipitates from extracts of HeLa S3 cells stimulated with TNF-α (20 ng/ml) for the indicated times were assayed for 1 hour by a standard immune complex kinase assay. Either GST–IκB-α (1–54) WT or the GST–IκB-α (1–54; S32T, S36T) mutant (S→T) (4 μg) was used as substrate. (Lower) IκB-α phosphorylation and degradation kinetics. HeLa cell extracts prepared as described in the upper panel were examined by protein immunoblotting for IκB-α degradation. IκB-α supershifting, a result of stimulus-dependent phosphorylation, is observed after 3 and 5 min of stimulation followed by the disappearance of IκB-α. (B) Stimulus-dependent activation of IKK signalsome is blocked by TPCK. Anti–MKP-1 immunoprecipitates from cell extracts of HeLa S3 cells stimulated for 7 min with TNF-α (20 ng/ml; lanes 2 and 6), IL-1 (10 ng/ml; lane 3), or PMA (50 ng/ml; lane 4), or treated for 30 min with TPCK (15 μM; lane 7) before stimulation with TNF-α, were examined for IKK signalsome activity. GST–IκB-α (1–54) WT (4 μg) was used as substrate. (C) The IKK signalsome phosphorylates Ser32 and Ser36 of IκB-α. (Upper) IκB-α (21–41) peptides that were unphosphorylated or that had been synthesized with P-Ser at position 32 or 36 were enzymatically phosphorylated by the IKK signalsome with [γ-32P]ATP. The unrelated c-Jun (56–70), c-Jun (65–79), and MKP-1 (349–366) peptides functioned as poor substrates for the IKK signalsome. (Lower) The same set of substrates described for the upper panel were subjected to enzymatic phosphorylation by JNK2 with [γ-32P]ATP used as a control. Specific peptide substrates used are indicated on top. Source of the kinase is indicated on the left. Molecular mass standards (in kilodaltons) are indicated on the right. (D) The IKK signalsome specifically phosphorylates Ser32 and Ser36 of the IκB-α protein and RelA in the context of a RelA–IκB-α complex. Anti–MKP-1 immunoprecipitates from cell extracts of HeLa S3 cells stimulated with TNF-α (20 ng/ml; 7 min) were examined for their ability to phosphorylate recombinant RelA–IκB-α complex containing either WT IκB-α (lane 3) or IκB-α (S32A, S36A) mutant (lane 4) protein. Specific substrates used are indicated on top. Positions of phosphorylated substrates are indicated by arrows on the left.

We established the substrate specificity of the IKK signalsome by using various peptides and recombinant proteins (8, 13). The kinase was capable of phosphorylating an IκB-α (21–41) peptide as well as two additional IκB-α (21–41) peptides, each bearing an unmodified Ser at position 32 or 36 and phosphoserine (P-Ser) at the other position (Fig. 2C). An IκB-α (21–41) peptide bearing Thr at both positions was phosphorylated <one-tenth as well as wild-type (WT) peptide, whereas an IκB-α (21–41) peptide bearing P-Ser at both positions was not phosphorylated at all (9). The IKK signalsome did not phosphorylate two c-Jun peptides containing Ser63and Ser73, respectively, or an MKP-1 peptide containing four Ser and three Thr (Fig. 2C). The latter peptides were substrates for JNK2. These experiments indicate that Ser32 and Ser36 were both specifically phosphorylated by the IKK signalsome.

The IKK signalsome phosphorylated WT IκB-α but not IκB-α (S32A, S36A) in the context of a physiological RelA–IκB-α complex (Fig. 2D). GST–IκB-β (1–44) was also phosphorylated, albeit with lower affinity; the Km for IκB-β was tenfold higher than the Km for IκB-α (9). The IKK signalsome also contained a strong RelA kinase activity that was distinct from the IκB kinase activity in that it was dissociated from the IKK signalsome by rigorous washing (9). No activity toward several other substrates was observed, including myelin basic protein (MBP), GST–activation transcription factor–2 (ATF2) (1–112), GST–c-Jun (1–79), GST–extracellular signal regulated kinase (ERK3), GST–Elk-1 (307–428), GST-p38, and a GST fusion protein containing the COOH-terminal region of IκB-α (242–314) (9).

We developed a two-step IKK signalsome purification method. Proteins from whole-cell lysates of TNF-α–stimulated HeLa cells were immunoprecipitated with anti–MKP-1. We eluted the IKK signalsome with an MKP-1 peptide and fractionated it further by anion-exchange chromatography (14). Fractions with IκB kinase activity were pooled and subjected to preparative SDS gel electrophoresis. Two prominent protein bands of 85 and 87 kD (designated IKK-1 and IKK-2, respectively, in Fig. 3B) correlated with the peak of IκB kinase activity. The 85- and 87-kD bands were excised, digested with trypsin, and analyzed by high mass accuracy matrixassisted laser deposition and ionization (MALDI) peptide mass mapping (15, 16). The 85-kD band was identified as CHUK (conserved helix-loop-helix ubiquitous kinase) (17), whereas the 87-kD band was not found in a comprehensive database. Three peptides derived from the 87-kD band were sequenced by nanoelectrospray tandem mass spectrometry (18) and found as identical matches to human expressed sequence tag (EST) clones (15) that were similar to human and mouse CHUK (17). Once the complete coding sequence of IKK-2 was obtained (19), all sequenced peptides (apart from two peptides derived from IKK-1) could be assigned to this protein (Fig. 3A).

Figure 3

Purification and cloning of IKK-1 and IKK-2. (A) Amino acid sequence comparison of human IKK-1 and IKK-2. Arrows, boundaries of the kinase domain; underlining, peptide sequences identified by nanoelectrospray mass spectrometry; boldface, amino acid identities conserved between IKK-1 and IKK-2; asterisks, amino acid identity to a region conserved among the MEK family of protein kinases demonstrated to be essential for stimulus-dependent activation; NH2-terminal boxed area, leucine zipper motif; COOH-terminal boxed area, helix-loop-helix domain; dashes, gaps inserted to optimize alignment. (B) Purified IKK signalsome fractions contain two prominent bands at 85 and 87 kDa. WCE was prepared from TNF-α–stimulated (20 ng/ml; 7-min induction) HeLa S3 cells (1.2 g of total protein). The IKK signalsome was immunoprecipitated from the HeLa S3 WCE with anti–MKP-1 antibodies, washed with buffer containing 3.5 M urea, and eluted overnight with excess MKP-1–specific peptide. Eluted IKK signalsome was subjected to Mono Q chromatography. Fractions containing active IKK signalsome activity were subjected to SDS-PAGE, and protein bands were visualized by a standard silver staining protocol. Peak IKK signalsome activity is associated with lanes 3, 4, and 5. Protein bands corresponding to IKK-1 and IKK-2 are indicated on the right. Molecular mass standards (kD) are indicated on the left.

Sequence analysis revealed that IKK-1 and IKK-2 are related protein serine kinases (51% identity) containing protein interaction motifs (Fig. 3A). Both contain the kinase domain at the NH2-terminus with a leucine zipper motif and a helix-loop-helix motif in the COOH-terminal region. Northern blot analysis indicated that mRNAs encoding IKK-2 were widely distributed in human tissues, with transcript sizes of about 4.5 and 6 kb (9). IKK-1 and IKK-2 mRNAs were expressed in Jurkat, HeLa, and HUVEC cell lines, and their amounts were not altered for up to 8 hours after stimulation of cells with TNF-α (HeLa and HUVEC) or antibody to CD28 plus PMA (Jurkat) (9).

Immunoprecipitates of epitope-tagged IKK-1 and IKK-2, expressed in rabbit reticulocyte lysates (RRLs) (20), phosphorylated IκB-α and IκB-β (Fig.4A). IKK-1 autophosphorylated (Fig. 4A), whereas a kinase-inactive version of IKK-1, in which the conserved Lys44 was mutated to Met (K44M), showed no autophosphorylation (9). IKK-2, although expressed in equivalent amounts in the lysates, showed very weak autophosphorylation. Immunoprecipitates of either IKK-1 or IKK-2 phosphorylate the IκB-α (21–41) peptide and two different IκB-α (21–41) peptides, each bearing an unmodified Ser at position 32 or 36 and P-Ser at the other position (Fig. 4B). IKK-1 and IKK-2, therefore, can independently phosphorylate both Ser32 and Ser36.

Figure 4

Expression of IKK-1 or IKK-2 generates IκB kinase activity. (A) Expression of IKK-1 or IKK-2 in RRLs generates IκB kinase activity. Either HA-tagged IKK-1 (lane 1) or Flag-tagged IKK-2 (lane 2) was translated in RRLs, immunoprecipitated, and examined for the ability to phosphorylate GST–IκB-α (1–54) WT and GST–IκB-β (1–44), as indicated by arrows. IKK-1 (lane 1) undergoes significant autophosphorylation in contrast to IKK-2 (lane 2), which autophosphorylates much more weakly. (B) IKK-1 or IKK-2 expressed in RRLs displays the ability to phosphorylate Ser32 and Ser36 of IκB-α. IκB-α (21–41) peptides that were unphosphorylated or that had been synthesized with P-Ser at position 32 or 36 were enzymatically phosphorylated by immunoprecipitates containing HA-tagged IKK-1 (Upper) or Flag-tagged IKK-2 (Lower) expressed in RRLs. The unrelated MKP-1 (349–366) peptide was included as a control. Specific peptide substrates used are indicated on top. (C) IKK-1 and IKK-2 display distinct modes of regulation when overexpressed in mammalian cells. HeLa cells were transiently transfected with either HA-tagged IKK-1 or Flag-tagged IKK-2 expression vectors. Thirty-six hours after transfection, cells were left unstimulated or were stimulated with TNF-α (20 ng/ml) for the indicated times. Activities of the tagged kinases were determined by immunokinase assays with GST–IκB-α (1–54) WT or GST–IκB-α (1–54; S32T, S36T) used as substrate. The transfected kinase is indicated on the left. (D) Endogenous IKK-1 and IKK-2 display activation kinetics identical to the IKK signalsome activity. Endogenous IKK-1 and IKK-2 were immunoprecipitated, with antibodies directed against IKK-1 and IKK-2, from HeLa WCEs prepared from cells that were left unstimulated or that were stimulated with TNF-α (20 ng/ml) for the indicated times. The immunocomplexes were then subjected to standard kinase assays with GST–IκB-α (1–54) WT or GST–IκB-β (1–44) WT as substrate, as indicated on the left. (E) A constitutively active version of IKK-2 is generated by (S177E, S181E) mutations in the activation loop resembling those of the MEK family kinases. (Upper) Flag-tagged IKK-2 WT (lane 1), (S177A, S181A) (lane 2), (S177E, S181E) (lane 3), and K44M (lane 4) were translated in RRLs, immunoprecipitated with anti-Flag, and examined for their ability to phosphorylate GST–IκB-α (1–54) WT. Phosphorylated GST–IκB-α (1–54) and autophosphorylated IKK-2 are indicated with an arrow on the left. IKK-2 mutants are indicated at the top. (Lower) An equal portion of each immunocomplex, as indicated in the upper panel, was subjected to Western blot analysis with anti-Flag to establish relative levels of IKK-2 protein expression.

Regulation of recombinant IKK-1 and IKK-2 activity, overexpressed in HeLa cells, appeared markedly different. Immunoprecipitates containing recombinant IKK-1 were inactive unless the cells were stimulated (Fig.4C). In contrast, IKK-2 immunoprecipitates yielded strong constitutively active IκB kinase activity in the absence of cell stimulation. However, immunoprecipitates containing the endogenous IKKs were inactive unless the cells had been stimulated (Fig. 4D). Interestingly, both IKK-1 and IKK-2 contain a canonical MAP kinase kinase (MAPKK) activation loop motif (Ser-Xaa-Xaa-Xaa-Ser, where Xaa is any amino acid) (Fig. 3A). Phosphorylation of both Ser residues is necessary for activation of MAPKK (21). We generated IKK-2 mutants in which Ser177 and Ser181 were mutated to Ala or Glu (S177A, S181A or S177E, S181E) to block or mimic, respectively, the effect of P-Ser. When expressed in RRL, IKK-2 (S177E, S181E) generated a highly active IκB-α kinase activity (Fig. 4E). The corresponding IKK-1 (S176E, S180E) mutation minimally enhanced kinase activity (9).

Both IKK proteins appear to have roles in NF-κB activation, although our data indicate that IKK-2 is more active. Immunocytochemical studies (22) showed that IKK-2 K44M and IKK-2 (S177A, S181A) mutants had no effect on subcellular localization of RelA in unstimulated HeLa cells. However, both mutants inhibited RelA nuclear translocation in TNF-α–stimulated cells (Fig.5A). The corresponding IKK-1 mutants, expressed at approximately equivalent amounts, had little inhibitory activity (Fig. 5A). The effects of the IKK-1 and IKK-2 mutants on NF-κB–dependent gene expression (Fig. 5B) paralleled those observed in the immunocytochemical studies (Fig. 5A). Both IKK-2 K44M and IKK-2 (S177A, S181A) inhibited TNF-α–stimulated NF-κB–mediated gene activation, whereas IKK-2 (S177E, S181E) induced activity in the absence of cell stimulation (Fig. 5B). Expression of IKK-1 mutants also perturbed NF-κB–mediated gene expression, although the effects were not as pronounced as for the IKK-2 mutants.

Figure 5

IKK-2 mediates an essential step in the NF-κB activation pathway. (A) IKK-2 mutants block stimulus-dependent RelA nuclear translocation. HeLa cells were transiently transfected with functionally equivalent mutants of either HA-tagged IKK-1 or Flag-tagged IKK-2 as indicated [HA-tagged IKK-1 WT, K44M (K→M), (S176A, S180A) (SS→AA), (S176E, S180E) (SS→EE); Flag-tagged IKK-2 WT, K44M (K→M), (S177A, S181A) (SS→AA), (S177E, S181E) (SS→EE)]. Thirty-six hours after transfection, cells were not stimulated or were stimulated with TNF-α (20 ng/ml) for 30 min (TNF-α). Cells were then subjected to immunocytochemical analysis with anti-HA or anti-Flag to visualize expression of IKK-1 and IKK-2, respectively. Anti-RelA were used to monitor stimulus-dependent nuclear translocation of RelA. Cellular immunofluorescence was analyzed for the presence of RelA nuclear staining as a function of either WT or mutant IKK-1 or IKK-2 expression. Data are presented as percentage of transfected cells expressing the indicated IKK protein that display clear RelA nuclear staining; >50 cells were scored per treatment. (B) Expression of IKK-2 mutants has a marked effect on NF-κB–dependent gene activation. HeLa cells were transiently cotransfected with a 3× NF-κB luciferase reporter vector and either an empty control vector or an IKK-2 expression vector as indicated. Thirty-six hours after transfection cells were not stimulated or were stimulated with TNF-α (20 ng/ml) for 5 hours before harvesting. Luciferase activities were determined and normalized on the basis of β-galactosidase activity. Average induction (fold) of luciferase activity was determined from a representative transfection experiment done in duplicate.

The IKK family of serine protein kinases are unique in that they contain both a leucine zipper and a helix-loop-helix interaction motif. We examined both in vitro and in vivo whether IKK-1 and IKK-2 form stable heterodimers. Influenza virus hemagglutinin (HA) epitope-tagged IKK-1 and Flag (IBI-Kodak) epitope-tagged IKK-2 were translated in RRLs, either alone or together, and then immunoprecipitated (Fig.6A). IKK-2 was present in IKK-1 immunoprecipitates and vice versa. Heterodimerization appears to be favored over homodimerization, because IKK-1 and IKK-2, when expressed separately and then combined, quantitatively form heterodimers (9). IKK-1 and IKK-2 also coimmunoprecipitate when coexpressed in HeLa cells (Fig. 6B). Removal of the leucine zipper and helix-loop-helix interaction motifs abrogated heterodimerization (9).

Figure 6

Interaction between IKK-1 and IKK-2. (A) IKK-1 and IKK-2 coprecipitate when translated in RRLs. HA-tagged IKK-1 and Flag-tagged IKK-2 were translated in vitro in RRLs either separately or in combination. The programmed translation mixture was then subjected to immunoprecipitation with the indicated antibody. Samples were subjected to SDS-PAGE and autoradiography. (B) IKK-1 and IKK-2 coprecipitate when coexpressed in HeLa cells. HeLa cells were transiently cotransfected with HA-tagged IKK-1 and Flag-tagged IKK-2. Thirty-six hours after transfection, cells were harvested and WCEs were prepared. Lysates were then immunoprecipitated with anti-flag or anti-APC (adenomatous polyposis coli). The immunocomplex was subjected to SDS-PAGE and protein immunoblotting with anti-HA or anti-Flag as indicated on the left.

The IκB kinase responsible for the initial and critical step of NF-κB activation has been the subject of intense interest. Many kinases have been proposed as candidates (23), but only the recently described CHUK IKK-α (24), which we refer to as IKK-1, has the characteristics expected of a cytokine-inducible IκB kinase. We have identified IKK-1 and the closely related kinase IKK-2 as interacting components of the IKK signalsome, a multiprotein signaling complex that regulates NF-κB activation in response to proinflammatory cytokines. Our results strongly suggest that IKK-1 and IKK-2 are functional kinases within the IKK signalsome that mediate IκB phosphorylation and NF-κB activation. As a protein complex containing multiple interacting components, including a RelA kinase, the IKK signalsome has the potential to integrate the diverse signaling pathways known to activate NF-κB in different cell types and channel them toward selective gene expression. Drugs that modulate the activation and function of the IKK signalsome are likely to have therapeutic value in inflammatory and neurodegenerative diseases as well as in cancer.

  • * To whom correspondence should be addressed. E-mail: fmercuri{at}


Stay Connected to Science

Navigate This Article