A cytokine-responsive IκB kinase that activates the transcription factor NF-κB

Abstract

Nuclear transcription factors of the NF-κB/Rel family are inhibited by IκB proteins, which inactivate NF-κB by trapping it in the cell cytoplasm. Phosphorylation of IκBs marks them out for destruction, thereby relieving their inhibitory effect on NF-κB. A cytokine-activated protein kinase complex, IKK (for IκB kinase), has now been purified that phosphorylates IκBs on the sites that trigger their degradation. A component of IKK was molecularly cloned and identified as a serine kinase. IKK turns out to be the long-sought-after protein kinase that mediates the critical regulatory step in NF-κB activation.

Main

Transcription factors of the NF-κB/Rel family are regulated through interactions with IκBs, which are inhibitory proteins^1,2. With the exception of mature B cells where they are consititutively nuclear, in all other cell types NF-κB dimers are kept in the cytoplasm through association with the IκBs, which mask their nuclear localization sequence^1,2. In response to diverse extracellular stimuli, including proinflammatory cytokines, viral infection, oxidants, phorbol esters and ultraviolet irradiation, the IκBs are rapidly phosphorylated at two serineswithin their amino-terminal regulatory domains. In two IκB proteins, site-directed mutagenesis^3,4,5,6,7 and biochemical mapping⁷ has identified these serines as S32 and S36 in IκBα and S19 and S23 in IκBβ. Phosphorylation at these sites triggers polyubiquitination of the IκBs^7,8,9 and, as with other proteins¹⁰, targets them for rapid degradation by the 26S proteasome^7,8. Proteasome inhibitors cause accumulation of IκB, which in stimulated cells is N-terminally phosphorylated, and inhibit NF-κB activation^11,12. Substitution of two lysines in IκBα, K20 and K21, one of which is conserved in IκBβ, with arginine residues decreases the rate of inducible IκBα ubiquitination and degradation without affecting its phosphorylation⁷.

As N-terminal phosphorylation of the IκBs occurs before and is necessary for their ubiquitination and degradation, it is likely to be the principal point of control through which diverse stimuli effect NF-κB activation¹³. However, it is not yet clear whether increased N-terminal phosphorylation of the IκBs is due to activation of an IκB kinase or inhibition of an IκB phosphatase^14,15. Part of this doubt is due to the fact that a cytokine-responsive protein kinase(s) phosphorylating IκBα at S32/36 or IκBβ at S19/23 has not been previously identified. In addition, okadaic acid, an inhibitor of protein-phosphatase 2A (PP2A), is a potent NF-κB activator and an inducer of IκB N-terminal phosphorylation⁸,¹⁴,¹⁵. In fact, an activity that phosphorylates IκBα at S32/36 was detected in okadaic acid-treated cell extracts and had an apparent relative molecular mass (M_r) of 700K (ref. 15). It was not established, however, whether this activity is stimulated by proinflammatory cytokines or other agonists and, if so, whether its activation kinetics correlate with those of IκB phosphorylation.

To understand better the signalling mechanism responsible for NF-κB activation, we tested cell extracts for the presence of a protein kinase activity that is activated by tumour-necrosis factor (TNF) and phosphorylates IκBα at S32/36. Such an activity was detected in extracts of TNF-treated HeLa cells and its substrate specificity and kinetics of activation correlated well with those of IκBα phosphorylation in living cells⁷. We purified this activity and determined a partial peptide sequence for one of its components. Molecular cloning and functional analysis identified this subunit as a protein kinase whose associated IκB kinase activity is rapidly stimulated by proinflammatory cytokines. This activity is inhibited upon dephosphorylation with PP2A. We demonstrate that this protein kinase, IKKα, is critical for NF-κB activation in response to proinflammatory cytokines.

Purification of an IκBα kinase

To investigate whether phosphorylation of IκBα at S32/36 in response to TNF is due to activation of an IκB kinase specific for these sites, we fractionated extracts of non-stimulated and TNF-stimulated HeLa cells on a Mono-Q column. Column fractions were tested for their ability to phosphorylate a glutathione S-transferase (GST)–IκBα fusion protein containing the N-terminal regulatory domain (residues 1 to 54). The specificity of the kinase and its physiological relevance were examined by using the mutant substrate GST–IκBα(1–54 TT), in which serines 32 and 36 are substituted with threonines⁷. In intact cells this substitution decreases the efficiency of IκBα phosphorylation, indicating that the relevant kinase strongly prefers serines over threonines⁷. As shown in Fig. 1a , TNF stimulation of HeLa cells resulted in rapid activation of a kinase activity that phosphorylates GST–IκBα(1–54). This activity did not phosphorylate GST–IκBα(1–54; TT). Fractionation of extracts of non-stimulated and TNF-stimulated HeLa cells by gel filtration revealed a TNF-stimulated IκBα(1–54) kinase activity that eluted with an apparent M_r of ∼900K (Fig. 1b). This activity was not detected when GST–IκBα(1–54; TT) was used as a substrate and also did not phosphorylate GST–IκBα(1–54; AA), in which serines 32 and 36 were substituted with alanines (Fig. 1c). This kinase activity was activated rapidly, peaking 5–10 min after TNF addition (data not shown).

Figure 1: Identification of a TNF-stimulated IκBα kinase activity.

a, Extracts of non-stimulated or TNF-stimulated HeLa cells were fractionated on a Mono-Q column. Top, elution profile of bulk protein. Bottom, elution of IκBα kinase activity assayed with GST–IκBα(1–54) as a substrate. Fold-stimulation of kinase activity in each fraction was determined by phosphorimaging and is indicated at the bottom. b, Extracts of TNF-stimulated and non-stimulated HeLa cells fractionated on a Superose-6 column. Top, phosphorylation of GST–IκBα(1–54) (arrow) by non-stimulated and TNF-stimulated extracts. Bottom, phosphorylation of GST–IκBα(1–54; TT) by TNF-stimulated cell extract. Positions of M_r markers are indicated at the top; the first lane in each panel is input. c, Substrate specificity of the most active fractions of each extract. GST–IκBα(1–54) (WT, GST–wild type)GST-IκBα(1–54; AA), and GST–IκBα(1–54; TT) were used as substrates. Coomassie blue(CB)-stained substrates and ³²P-labelling (KA, kinase assay) are shown.

We purified this kinase activity from large-scale cultures of HeLa cells stimulated with TNF. Cytosolic extracts (S-100 fraction) were chromatographed on Q-Sepharose and the active fractions were pooled and chromatographed through a γ-phosphate-linked ATP–Sepharose¹⁶ affinity column. Bound proteins were eluted with ATP and the active fractions pooled and chromatographed on a Superose-6-gel-filtration column. This procedure resulted in ∼25,000-fold enrichment of IκBα kinase activity (Table 1). The purified kinase, which we call IκB kinase (IKK), has the same substrate specificity as the TNF-stimulated activity already described (Fig. 2c). To examine the polypeptide composition of IKK, we analysed the Superose-6 column fractions by SDS–polyacrylamide gel electrophoresis (SDS–PAGE) and silver staining. Only two polypeptides, of 85K and 87K, clearly co-eluted with IKK activity at this stage (Fig. 2a). As even the most active fraction contained a number of polypeptides, we applied the pooled Superose-6 IKK fractions to an affinity column of a GST–IκBα(1–54; AA)_{8 ×} fusion protein covalently linked to agarose. After extensive washing, the column was developed with a NaCl gradient. This resulted in a further fourfold enrichment of IKK activity (Table 1). The composition of the material retained on the affinity column is shown in Fig. 2b . Although this fraction contains several polypeptides, we consistently detected enrichment of the 85K and 87K band, as well as of the 64K band. It is not clear yet whether the other bands represent additional IKK components or contaminants. Gel-filtration analysis of this material indicated that it continued to elute as a very large (∼900K) complex.

Table 1 Purification of IKK activity

Figure 2: Purification and substrate specificity of IκBα kinase (IKK).

a, Cytoplasmic extracts from TNF-stimulated HeLa cells were purified as described. Top, active Superose-6 fractions analysed by SDS–PAGE and silver staining. Fraction numbers are indicated at the top, M_r markers on the right. Arrows, bands that coelute with kinase activity. Bottom, kinase assay (KA) of each fraction with relative activity (RA) indicated underneath. b, Fraction 35 was further purified by substrate-affinity chromatography using GST–IκBα(1–54; AA)_{8 ×} covalently bound to Sepharose. I, input; FT, flow through; NaCl, eluate. The 85K and 87K bands are indicated. Bottom, kinase assay (KA). c, Purified IKK was tested for its ability to phosphorylate GST–IκBα(1–54)(WT), GST–IκBα(1–54; AA), GST–IκBα(1–54; TT), full-length IκBα (α-FL), GST–IκBβ(1–44)(WT), GST–IκBβ(1–44; AA), and full-length IκBβ (β-FL). Kinase assay (KA, top) and Coomassie blue (CB, bottom)-stained proteins are shown.

Cloning of an IKK protein-kinase subunit

We purified the 85K polypeptide associated with IKK to homogeneity by preparative SDS–PAGE and subjected it to microsequencing. In addition to using oligonucleotide primers based on the obtained sequence, we searched Genebank for related sequences and found that both peptides were contained within a partial sequence of a putative human serine/threonine kinase with unknown function named CHUK¹⁷. Polymerase chain reaction (PCR) and library screening were used to isolate the complete complementary DNA. Nucleotide sequencing revealed the presence of an open reading frame for a 744-amino-acid polypeptide (Fig. 3; Genebank accession number AF009225). The N-terminal half of the predicted polypeptide contains a 301-amino-acid protein-kinase domain, whereas its C-terminal half, as previously reported¹⁷, contains several protein-interaction motifs, including a leucine zipper and a helix–turn–helix motif. In the light of its function, we renamed the protein encoded by this cDNA as IKKα (for α subunit of IKK).

Figure 3: The predicted amino-acid sequence of IKKα: the two peptides whose sequence was determined are overlined.

The kinase domain (KD), leucine zipper (LZ) and helix–loop–helix (HLH) are indicated. Peptide sequences used for primer design are indicated by underlining arrows; the sequences of the three primers are shown.

Cell-free translation of in vitro generated IKKα transcripts resulted in production of a polypeptide of the expected size (Fig. 4a; cell-free-translated IKKα migrates more slowly than the p85 IKK subunit owing to the presence of a haemagglutinin(HA) epitope). We tested this protein for IκBα kinase activity. HA–IKKα protein was immunoprecipitated with HA antibody, extensively washed in urea at concentrations of up to 3 M, and the immune complexes incubated with various GST–IκB proteins or GST–c-Jun in the presence of [γ-³²]ATP. This resulted in the efficient phosphorylation of either GST–IκBα(1–54) or full-length GST–IκBα (Fig. 4b). Phosphorylation of either GST–IκBβ(1–44) or full-length GST–IκBβ was less efficient, and GST–IκBα(1–54; TT), GST–IκBα(1–54; AA), GST–IκBβ(1–44; AA) or GST–c-Jun(1–79) were not phosphorylated. This profile of substrate specificity is identical to that of native IKK purified from TNF-treated HeLa cells (Fig. 2c).

Figure 4: Cell-free translation of IKKα generates IκB kinase activity.

a, Cell-free-translated HA-IKKα labelled with ³⁵S-methionine was separated by SDS–PAGE next to affinity-purified IKK stained with silver. b, Mock-programmed (V), HA-IKKα- or HA–JNK1-programed reticulocyte lysates were immunoprecipitated with HA antibody and, after washing with 3M urea, immune complex kinase activity was assayed with GST–IκBα(1–54) as substrate. ³⁵S- and ³²P-labelled proteins were detected by autoradiography. c, After translation of HA–IKKα and immunoprecipitation, the immune complexes were incubated with either no exogenous substrate (−) or the indicated substrates in kinase buffer and [γ-³²P]ATP. Top, kinase assay (KA) revealed by autoradiography; bottom, Coomassie blue (CB)-stained substrates.

Expression and activation of IKKα

The most efficient NF-κB activators are the proinflammatory cytokines interleukin (IL)-1 and TNF¹³,¹⁸. These cytokines are also efficient inducers of IκBα phosphorylation and degradation⁷. We examined the effect of TNF and IL-1 on IKKα-associated kinase activity. HA–IKKα was transiently expressed in HeLa cells. As a control, we transiently transfected HeLa cells with a HA–JNK1 expression vector. After 36 h, these cells were stimulated with TNF or IL-1 for 5 min and the HA-tagged protein kinases immunoprecipitated. HA–IKKα immunoprecipitates from TNF- or IL-1-stimulated cells phosphorylated GST–IκBα(1–54) but did not phosphorylate GST–c-Jun(1–79) (Fig. 5a). By contrast, HA–JNK1 phosphorylated GST–c-Jun(1–79) but not GST–IκBα(1–54). The extent of IKKα activation by TNF or IL-1 was similar to the extent of JNK activation by these cytokines. Time-course analysis indicated that stimulation of IKKα activity occurred rapidly (Fig.5b). Maximal IκBα kinase activity was reached within 5 min of TNF or IL-1 addition. After 15–30 min, this activity rapidly declined. IKKα activity was also moderately stimulated by the phorbol ester 12-O-tetradecanoylphorbol-13-acetate acetate (TPA; data not shown).

Figure 5: Expression and cytokine activation of IKK in mammalian cells.

a, HeLa cells were transiently transfected with either HA–JNK1 or HA–IKKα expression vectors. After 36 h, cells were incubated with either control medium (−), TNF or IL-1 for 5 min and the activities of the HA-tagged kinases were determined by immune-complex kinase assays with GST–c-Jun(1–79) or GST–IκBα(1–54) as substrates. The recovery of HA–JNK1 and HA-IKKα was determined by immunoblotting. b, HeLa cells were transfected with HA–IKKα expression vector and were stimulated after 36 h with either TNF or IL-1 for the indicated times, at which point cells were lysed and IKK activity determined by immune-complex kinase assay.

We investigated the substrate specificity of HA–IKKα immune complexes isolated from TNF-stimulated cells and found it to be identical to that of purified IKK (data not shown). Phosphopeptide mapping on Tris–Tricine gels⁷ revealed that phosphorylation of GST–IκBα(1–54) occurred on the peptide that contains S32 and S36 (Fig. 6). The alanine-substitution mutant GST–IκBα(1–54; AA) was not phosphorylated at all (data not shown; but see Figs 2c and 4c). The use of single-substitution mutants, IκBα(A32) or IκBα(A36), indicated that both purified IKK and HA-IKKα immune complexes phosphorylated both S32 and S36 with comparable efficiency (Fig. 6b , c). However, phosphopeptide mapping of full-length IκBα phosphorylated by either purified IKK or HA-IKKα immune complexes revealed that, in addition to the peptide containing S32 and S36, a smaller amount of ³²P was incorporated into a peptide spanning amino acids 264 to 314 (Fig. 6a , c). When full-length IκBα(A32/36) was used as a substrate, all of the ³²P was incorporated into that C-terminal peptide. Phosphoaminoacid analysis indicated that phosphorylation occurred at serine residues. Previously we have localized the constitutive phosphorylation sites of IκBα to the same peptide near its C terminus⁷ (J.A.D., unpublished results). It is not yet clear, however, whether IKK phosphorylates exactly the same sites.

Figure 6: Mapping of IKK phosphorylation sites on IκBα.

a, GST–IκBα(1–54), full-length (FL) WT or A32/36 IκBα were phosphorylated by either purified IKK or HA–IKKα immune complexes isolated from TNF-stimulated cells. Equal counts of each protein were digested with trypsin and the resulting digests separated on Tris–Tricine gels together with M_r markers. The ∼0.9K peptide corresponds to amino acids 30–38 and the ∼6K peptide corresponds to amino acids 263–314 (ref. 7). The location of these tryptic peptides is shown at the top. b, Extracts were prepared from HeLa cells stably expressing HA–IKKα that were either non-stimulated or stimulated with TNF for 10 min and immunoprecipitated with either nonspecific (NS) or HA-antibody. The immune complexes were used to phosphorylate full-length WT IκBα, IκBα(A32/36), IκBα(A32) or IκBα(A36). The same proteins were also phosphorylated by purified IKK. Equal amounts of each substrate were analysed by SDS–PAGE and autoradiography. c, The proteins in b were subjected to tryptic phosphopeptide mapping and the digests resolved on a Tris–Tricine gel.

IKKα is involved in NF-κB activation

To investigate further the role of IKKα in IκB phosphorylation in intact cells, we generated a stably transfected pool of HT-29 cells expressing HA-IKKα. Previous studies indicated that TNF induces partial IκBα degradation in these cells with slower kinetics than in other cell lines, especially HeLa (J.A.D., unpublished results). Expression of HA-IKKα in HT-29 cells markedly accelerated IκBα phosphorylation (indicated by the appearance of a slower migrating form⁷) and degradation in response to TNF (Fig. 7). The kinetics of IκBα degradation correlated with the kinetics of IKK activation.

Figure 7: Expression of IKKα accelerates IκBα degradation.

Pools of HT-29 cells stably expressing HA–IKKα and the parental cells (HT-29) were stimulated with TNF for the indicated times at which point the cells were collected and lysed. Cell lysates were tested by immunoblotting for IκBα degradation⁷ and HA–IKKα expression, as well as IKK activity of HA–IKKα immune complexes (KA). The basal and the hyperphosphorylated forms of IκBα are indicated, as is the HA–IKKα band. NS, nonspecific. HA imunoprecipitates of the parental cells do not contain IKK activity (not shown).

We tested the effect of raised or reduced IKKα expression on NF-κB activation by transient transfection. A 2 × NF-κB–Luc reporter¹² was cotransfected with an empty expression vector, an HA–IKKα expression vector, or an antisense (AS) IKKα vector into HeLa cells. Expression of HA–IKKα potentiated basal reporter gene activity and its stimulation by TNF or IL-1 (Fig. 8a). By contrast, cotransfection with AS-IKKα blocked induction of 2 × NF-κB reporter activity by IL-1 or TNF. In other experiments, AS-IKKα inhibited induction of 2 × NF-κB–Luc by TPA (data not shown). The inhibition was specific as it was not seen upon cotransfection of either JNKK1 or MKK3 antisense vectors. In addition, the inhibitory effect of antisense IKKα was reversed upon cotransfection of the HA–IKKα vector (Fig. 8b). Furthermore, immunoblot analysis indicated that AS-IKKα inhibited expression of HA–IKKα but not HA–JNK1 (data not shown). The stimulatory effect of HA–IKKα on NF-κB reporter activity was specific as it was not detected using a β-actin–LacZ reporter (Fig. 8c).

Figure 8: IKKα and NF-κB activation.

a, HeLa cells were transiently cotransfected with 2 × NF-κB–Luc reporter¹² together with empty expression vector, HA–IKKα expression vector, antisense (AS) IKKα vector (IKKα cDNA cloned in 3′ to 5′ orientation downstream of the β-actin promoter), antisense JNKK1 vector or antisense MKK3 vector. After 36 h, cultures were divided and either left untreated or incubated with either TNF or IL-1. After 3 h, luciferase activity was determined. The average fold-induction of luciferase activity in 3 separate experiments is shown. b, Cells were cotransfected with empty expression vector, HA-IKKα or antisense IKKα vectors, alone or in combination. After 36 h, cells were either stimulated or not with TNF for 3 h and luciferase activity determined. c, Cells were cotransfected with the indicated reporters together with the indicated amounts of HA–IKKα expression vector. After 36 h, cells were stimulated or not with TNF; after 3 h, luciferase and β-galactosidase activities were determined. Results are also shown from a β-actin–Lac Z reporter.

IKK is inactivated by PP2A

The identification of a cytokine-activated IκB kinase suggests that the activation of NF-κB and induction of IκBα phosphorylation by the PP2A-inhibitor okadaic acid¹⁴,¹⁵ could be due to negative regulation of IKK itself by PP2A. Such an interpretation is consistent with the rapid inactivation of IKK activity following its stimulation by TNF or IL-1. To test this possibility, we performed the experiment shown in Fig. 9a . Whereas okadaic acid did not affect IκBα phosphorylation by IKK purified from TNF-treated HeLa cells, inclusion of PP2A in the reaction or preincubation of IKK with PP2A diminished IκBα phosphorylation. This effect was inhibited by okadaic acid. To determine the target for PP2A action, we preincubated purified IKK enzyme with PP2A and then performed the kinase reaction in the presence of okadaic acid, which caused a large decrease in IκBα phosphorylation. When PP2A was added during the kinase reaction together with okadaic acid, no effect on IκBα phosphorylation was seen. Hence, IKK activity is sensitive to PP2A. We also found that, once phosphorylated by IKK, IκBα was not dephosphorylated by PP2A (Fig. 9b).

Figure 9: IKK is sensitive to PP2A.

a, IKK was preincubated with or without PP2A catalytic subunit (gift from G. Walter) at 30 °C for 1 h and then assayed for GST–IκBα(1–54) phosphorylation. Where indicated, okadaic acid (OA) was added. b, GST–IκBα(1–54) was phosphorylated by IKK and the kinase was heat-inactivated (h.i.). The phosphorylated substrate was then incubated with PP2A with or without OA as indicated (lanes 1, 2). Lane 3, control kinase reaction; lane 4; PP2A was present throughout the reaction. c, HA–IKKα was immunoprecipitated from transiently transfected HeLa cells incubated in either control medium (C) or in the presence of TNF for the last 5 min (T). The immune complexes were preincubated in the absence (−) or present (+) of PP2A for 1 h and then tested for GST–IκBα(1–54) phosphorylation. As a control, HA immune complexes were also isolated from non-transfected cells (first four lanes). Last two lanes, PP2A, sensitivity of purified IKK. d, HeLa cells stably expressing HA–IKKα were incubated with 50 nM okadaic acid for the indicated times, then cells were lysed and the IKK activity assayed with GST–IκBα(1–54) as substrate (top). Degradation of IκBα in these cells was determined by immunoblotting (bottom).

The IκB kinase activity associated with IKKα is also sensitive to PP2A. HA-IKKα immune complexes from transiently transfected non-stimulated or TNF-stimulated cells were treated or not with PP2A and tested for IκBα phosphorylation in the absence or presence of okadaic acid. Like the purified enzyme, the activity of HA–IKKα was sensitive to PP2A (Fig. 9c). Consistent with these results, we find that incubation of HeLa or HT-29 cells, stably expressing HA–IKKα, with okadaic acid caused a slow but considerable increase in IKK activity (Fig. 9d). In addition, cotransfection with the AS–IKKα vector inhibited induction of 2 × NFκB–Luc by okadaic acid (data not shown).

Discussion

The proinflammatory cytokines TNF and IL-1 exert many of their effects through activation of transcription factors AP-1 and NF-κB, which mediate induction of genes encoding other cytokines and various proteins involved in the inflammatory response^18,19,20. We understand the immediate events involved in TNF (refs 21,22,23) and IL-1 (refs 24, 25) signalling and how they lead to activation of the Jun N-terminal kinase (JNK) and p38 MAPK cascades that mediate induction of AP-1 activity²⁶ (Z. G. Liu, unpublished results), but not how these early signalling events lead to NF-κB activation. Although IκB phosphorylation is critical for cytokine-mediated NF-κB activation, the protein kinase(s) responsible was previously unidentified. It was even suggested that increased IκB phosphorylation following cytokine treatment was due to inhibition of a phosphatase. We have now described the purification and molecular cloning of a component of the cytokine-activated protein kinase complex IKK which is responsible for inducible IκB phosphorylation.

Several lines of evidence indicate that IKK is the critical protein kinase mediating NF-κB activation by TNF or IL-1. First, IKK activity is rapidly stimulated by TNF, IL-1 or TPA and its kinetics of activation match those of IκBα phosphorylation and degradation in intact cells. Second, IKK phosphorylates IκBα and β on the same residues, S32/36 and S19/23, respectively, that are phosphorylated in intact cells in response to TPA, IL-1 or TNF⁷. Phosphorylation of these serines triggers the polyubiquitination and degradation of IκB^3,4,5,6,7,8. Third, IKK does not phosphorylate mutants of IκBα in which these serines are replaced with threonines. Such mutants are poorly phosphorylated in intact cells and are refractory to cytokine-induced degradation⁷. Fourth, elevated expression of IKKα, the subunit of IKK that we have cloned, accelerates IκBα degradation and potentiates NF-κB activation, whereas expression of an antisense IKKα construct blocks NF-κB activation by TNF or IL-1. Fifth, IKK activity is sensitive to PP2A, whereas the specific PP2A inhibitor, okadaic acid, activates NF-κB¹⁴,¹⁵ and IKK. On the other hand, PP2A does not dephosphorylate IκBα previously phosphorylated by IKK.

IKK purifies as a very large and stable species composed of several polypeptides. The exact polypeptide composition of IKK remains to be determined and the protein kinase we have cloned, IKKα, is only one subunit of that complex. Although the size of the IKK complex is similar (but not identical) to that of a previously reported IκB kinase¹⁵, the two activities may not be related. The activity of the enzyme studied by Chen et al.¹⁵ depends on its ubiquitination, but we find no evidence that ubiquitination is required for IKK activation. Rather, the critical modification for IKK activation, based on its sensitivity to PP2A, appears to be phosphorylation. In addition, the previously described IκB kinase activity appears to be constitutive¹⁵, whereas IKK activity is rapidly stimulated by cytokines. Nevertheless, IKK isolated from unstimulated cells has a basal activity, amounting to 5–10% of that of the activated enzyme. Indeed, overexpression of IKKα can enhance NF-κB activity even in unstimulated cells. In addition, the upstream inputs that lead to IKK activation are weakly active in unstimulated cells. Inhibition of PP2A, the phosphatase that inactivates IKK, results in slow but considerable IKK activation. The nature of the input responsible for basal IKK activity remains to be identified. It has been shown²⁷ that the ubiquitination-dependent IκB kinase can be further activated in vitro by MEKK1; however, expression of a catalytically inactive MEKK1 mutant inhibits JNK activation without affecting NF-κB activation²⁶ and so far we have been unable to activate IKK in vitro with MEKK1 (unpublished results).

As IKKα is a serine/threonine kinase by sequence and its cell-free translation produces an IκB kinase activity with the same substrate specificity as native IKK, it is likely that it is a catalytic subunit of this complex. IKKα, which is essentially identical in sequence to the putative serine/threonine kinase CHUK, whose function and mechanism of regulation were previously unknown¹⁷, contains protein-interaction motifs in its C-terminal half, including a leucine zipper and a helix–turn–helix. These motifs could mediate interaction of IKKα with other subunits of the IKK complex. Although the IκBs are not part of this complex, IKK binds to an IκBα-affinity column. The IKK subunit responsible for this interaction remains to be identified. Previously, by studying the interaction of JNK with c-Jun, we demonstrated that docking of a signal-regulated protein kinase to its substrate is important for rapid and specific phosphorylation²⁸.

Although in Jurkat cells TNF induces the degradation of IκBα but not IκBβ²⁹, in several other cell lines we found that the main difference between the two IκBs was in the rate of their degradation: IκBα is degraded faster than IκBβ⁷. This could be caused by a limiting amount of a protein kinase with a higher affinity for IκBα than for IκBβ. IKK exhibits this property in vitro, phosphorylating IκBα more efficiently than IκBβ. Also, if IκBβ degradation contributes to NF-κB activation during a 3-hour stimulation in HeLa cells, the fact that antisense IKKα RNA blocks NF-κB activation indicates that IKK may also be an IκBβ kinase. But it remains to be seen whether overexpression or activation of IKK accelerates IκBβ phosphorylation and degradation.

Our results establish IKK as the long-sought-after protein kinase that phosphorylates at least IκBα in response to proinflammatory cytokines. The identification of IKK and the molecular cloning of one of its subunits will lead to a better understanding of the signalling pathways originating at the TNF and IL-1 receptors that depend on recruitment of TRAF mediators^{21,22,23,24,25} to elicit activation of NF-κB and the inflammatory gene-induction response.

Methods

Protein purification. Suspension cultures of HeLa S3 cells were stimulated with 20 ng ml⁻¹TNF (R & D Systems) for 5 min at 5 × 10⁶ cells ml⁻¹. Cells were collected and lysed in buffer A (20 mM Tris-Cl, 20 mM NaF, 20 mM β-glycerophosphate, 19 mM PNPP, 500 µM Na₃ VO₄, 2.5 mM metabisulphite, 5 mM benzamidine, 1 mM EDTA, 0.5 mM EGTA, 1 mM PMSF, 10% glycerol, pH 7.6) supplemented with 20 µg ml⁻¹aprotonin, 2.5 µg ml⁻¹ leupeptin, 8.3 µg ml⁻¹ bestatin, 1.7 µg ml⁻¹ pepstatin and 0.05% NP-40, by using a glass Dounce homogenizer. After centrifugation at 12,000 r.p.m. for 20 min in a Beckman SS34 rotor, the supernatant was recentrifuged at 38,000 r.p.m. for 80 min in a Beckman Ti 50.1 rotor, all at 4 °C. The supernatant (S-100) was flash-frozen and stored at −80 °C. Thawed S-100 fractions were purified over a 56 ml Q-Sepharose FF column equilibrated with buffer A and 0.1% Brij-35 (buffer B). After washing with buffer B plus 100 mM NaCl, the column was eluted with a linear 100–300 mM NaCl gradient. Fractions containing peak IKK activity were pooled, diluted 1:4 and applied to a 5-ml Hi Trap-Q column (Pharmacia). After washing with buffer B, the column was eluted in buffer B plus 300 mM NaCl. IKK-containing fractions were pooled and diluted1:1 with ATP-column buffer¹⁶. The material was passed 4 times over a 4-ml γ-phosphate-linked ATP-affinity column¹⁶. After washing with 10 ml ATP column buffer, 0.05% Brij-35, and with 10 ml ATP-column buffer, 0.05% Brij-35, 250 mM NaCl, the column was eluted with ATP-column buffer, 0.05% Brij-35, 250 mM NaCl, 10 mM ATP. After 1:3 dilution with buffer B, the eluate was applied to a 1-ml Hi-Trap-Q column. The column was eluted with buffer B plus 300 mM NaCl. IKK-containing fractions were pooled, concentrated and applied to a Superose-6 column, equilibrated and eluted in the same buffer. IKK-containing fractions were pooled and applied to an affinity column of a GST–IκBα(1–54; AA)_{8 ×} (eight repeats of the N-terminal domain) covalently linked to Sepharose after 1:4 dilution in buffer A. After washing in buffer A the column was eluted with a NaCl gradient.

IKK assay, immunoprocipitation and immunoblotting. Kinase activity was assayed in 20 mM HEPES, 20 mM β-glycerophosphate, 10 mM MgCl₂, 10 mM PNPP, 100 µM Na₃ VO₄, 2 mM DTT, 20 µM ATP, 10 µg ml⁻¹ approtonin, 50–200 mM NaCl, pH 7.5, and (1–10 µCi)[γ-³²P]ATP at 30 °C for 30 min. IκB-substrate proteins were expressed and purified from E. coli⁷. HA–IKKα immune complexes were isolated as described³⁰ and washed in kinase buffer containing 3M urea before determining kinase activity. Immunoblotting was done as described⁷.

Peptide sequencing and mapping. Polypeptides to be sequenced were purified by preparative SDS–PAGE. The 85K IKKα band was digested in situ with endoprotease Lys-C. The resultant peptides were eluted and purified by reverse-phase chromatography on an ABI 173 microblotter. Their sequence was determined on an ABI Procise Protein MicroSequencer. Phosphopeptide mapping on Tris–Tricine gels and amino-acid analysis were done as described⁷.

Plasmids, cell culture and transfections. The various expression vectors were constructed using standard recombinant DNA procedures. The β-actin promoter³¹ was used to drive expression of both HA–IKKα and antisense IKKα. Transient transfections were done as described²⁶,²⁸. Stably transfected cell lines were generated as described⁷,³¹ and identified by immunoblot screening of individual clones or pools of clones. Luciferase and β-galactosidase assays have also been described¹²,²⁶.