The activity of transcription factor NFκB is regulated by its subcellular localization. In most cell types, NFκB is sequestered in the cytoplasm due to binding of the inhibitory protein IκBα. Stimulation of cells with a wide variety of agents results in degradation of IκBα, which allows translocation of NFκB to the nucleus. Degradation of IκBα is triggered by phosphorylation of two serine residues, i.e. Ser32 and Ser36, by as yet unknown kinases. Here we report that the mitogen‐activated 90 kDa ribosomal S6 kinase (p90rsk1) is an IκBα kinase. p90rsk1 phosphorylates IκBα at Ser32 and it physically associates with IκBα in vivo. Moreover, when the function of p90rsk1 is impaired by expression of a dominant‐negative mutant, IκBα degradation in response to mitogenic stimuli, e.g. 12‐O‐tetradecanoylphorbol 13‐acetate (TPA), is inhibited. Finally, NFκB cannot be activated by TPA in cell lines that have low levels of p90rsk1. We conclude that p90rsk1 is an essential kinase required for phosphorylation and subsequent degradation of IκBα in response to mitogens.
Mammalian transcription factor NFκB is a key regulator of a broad range of genes involved in inflammatory processes, growth control and apoptosis (reviewed in Grilli et al., 1993; Baeuerle and Henkel, 1994; Beg and Baltimore, 1996; Liu et al., 1996; Wu et al., 1996). NFκB is a heterodimer that is composed of a 50 and a 65 kDa subunit. These subunits belong to a larger group of proteins, generally referred to as the family of Rel‐related proteins. The members of the family of Rel‐related proteins share sequence homology with the product of the v‐rel oncogene from the avian reticuloendotheliosis virus strain T (Stephens et al., 1983). On the basis of their structure, they can be divided into two subclasses. The first consists of p65‐RelA, relB and c‐Rel which are synthesized as mature proteins and have sequence homology within the N‐terminal 300 amino acids, the so‐called Rel homology domain (RHD) (reviewed in Verma et al., 1995). The RHD is responsible for DNA binding, dimerization and nuclear localization of the dimers (Logéat et al., 1991; Bressler et al., 1993). In addition, these proteins contain non‐conserved transactivation domains. The second class of Rel‐related proteins is composed of p50‐NFκB1 and p52‐NFκB2. These proteins are produced as precursor proteins p105‐NFκB1 and p100‐NFκB2, respectively (reviewed in Verma et al., 1995). In addition to a highly homologous RHD, these precursor proteins have a C‐terminal ankyrin repeat domain (ARD). The ARD is removed by ubiquitin‐dependent proteolytic processing in the 26S proteasome, giving rise to mature p50‐NFκB1 and p52‐NFκB2 (Fan and Maniatis, 1991; Palombella et al., 1994; Orian et al., 1995). With the exception of relB (which only forms heterodimers with p50‐NFκB1 or p52‐NFκB2), the members of both subclasses can form all possible homo‐ and heterodimers, generally referred to as Rel–NFκB.
Two types of inactive Rel–NFκB complexes are found in the cytoplasm of unstimulated cells. The first is formed by Rel dimers that are bound to the inhibitory proteins IκBα, IκBβ, IκBγ or Bcl‐3 (reviewed in Verma et al., 1995). These inhibitory proteins have ARDs consisting of 5–7 ankyrin repeats, which are required for binding to the RHD of Rel–NFκB. Binding of IκB to Rel–NFκB masks the nuclear localization signal (NLS) of the latter and, thus, causes cytoplasmic retention of the ternary complex (Kerr et al., 1991; Beg et al., 1992; Franzoso et al., 1992; Hatada et al., 1992; Inoue et al., 1992; Wulckzyn et al., 1992; Naumann et al., 1993). A second type of inactive complex is formed by the unprocessed precursor proteins p105‐NFκB1 and p100‐NFκB2 (Capobianco et al., 1992; Liou et al., 1992; Rice et al., 1992; Hatada et al., 1993; Mercurio et al., 1993; Naumann et al., 1993; Scheinmann et al., 1993; Dobrzanski et al., 1994). These proteins dimerize with mature Rel‐related proteins, giving rise to inactive heterodimers. The ARD in the C‐terminus of p105‐NFκB1 and p100‐NFκB2 is functionally homologous to IκB. It masks the NLS of the dimer and retains it in the cytoplasm.
Inactive cytoplasmic Rel–NFκB can be activated by stimulation of cells with a broad range of NFκB‐inducing agents (reviewed in Grilli et al., 1993; Baeuerle and Henkel, 1994; Israël, 1995), including mitogens (e.g. the phorbol ester 12‐O‐tetradecanoylphorbol 13‐acetate; TPA) and stress factors (e.g. tumour necrosis factor‐α; TNF‐α). Stimulation of cells with NFκB‐activating agents leads to phosphorylation of IκB, which is followed rapidly by its ubiquitin‐dependent degradation in the 26S proteasome (Brown et al., 1993; Henkel et al., 1993; Mellits et al., 1993; Rice and Ernst, 1993; Finco et al., 1994; Miyamoto et al., 1994; Traenckner et al., 1994; Chen et al., 1995; DiDonato et al., 1995; Lin et al., 1995). In addition, p105‐NFκB1 and p100‐NFκB2 are phosphorylated and processed (Mellits et al., 1993; Mercurio et al., 1993; Naumann and Scheidereit, 1994; MacKichan et al., 1996). Both the degradation of IκBα and the processing of p105‐NFκB1 result in unmasking of the NLS, which allows nuclear translocation of Rel–NFκB and activation of specific genetic programs. Much research has focused on the activation of the NFκB (the p50‐NFκB1–p65‐RelA dimer) bound to the inhibitory protein IκBα. Ligand‐induced phosphorylation of IκBα occurs at Ser32 and Ser36, giving rise to a phosphorylated form of IκBα which has a reduced mobility in SDS–PAGE as compared with non‐induced IκBα (Brockman et al., 1995; Brown et al., 1995; Traenckner et al., 1995; Whiteside et al., 1995; DiDonato et al., 1996). Mutation of both Ser32 and Ser36 completely inhibits ligand‐induced phosphorylation of IκBα, whereas mutation of one of these residues inhibits ligand‐induced IκBα phosphorylation only partially, yielding a phosphorylated form of IκBα with an electrophoretic mobility intermediate between those of non‐induced and ligand‐induced phosphorylated wild‐type IκBα (Traenckner et al., 1995; Whiteside et al., 1995). Furthermore, mutation of one or both of these serines renders IκBα resistant to ubiquitination and subsequent degradation in the 26S proteasome (Chen et al., 1995; DiDonato et al., 1996; Roff et al., 1996), showing that phosphorylation of both serines is an absolute requirement for the induction of IκBα degradation. However, the identity of the kinase or kinases that phosphorylate IκBα at these serine residues is not known.
The mitogen‐activated 90 kDa ribosomal S6 kinase (p90rsk1) is a serine/threonine kinase involved in the transduction of signals which are induced by stimuli that activate the Ras/mitogen‐activated protein (MAP) kinase cascade (Blenis, 1993). It is phosphorylated and subsequently activated by the extracellularly regulated kinases (ERKs) p44erk1 and p42erk2 but is also capable of autophosphorylation (Sturgill et al., 1988; Chung et al., 1991; Grove et al., 1993). To date, several isoforms of p90rsk1 have been described in mouse and human (Alcorta et al., 1989; Moller et al., 1994; Zhao et al., 1995). Originally, p90rsk1 was identified as a kinase that phosphorylates the S6 protein of the ribosomal 40S subunit in vitro (Erikson and Maller, 1986). However, later it became apparent that, instead of p90rsk1, the non‐related 70 kDa ribosomal S6 kinase (p70s6k) is responsible for mitogen‐induced phosphorylation of S6 in vivo (Chung et al., 1992; Price et al., 1992; Thomas, 1993). Interestingly, mitogen‐induced activation of p90rsk1 parallels the induction of immediate early genes, suggesting that p90rsk1 is involved in the regulation of some of these early events. Thus far, it has been shown that p90rsk1 phosphorylates the early transcription factors c‐Fos (Chen et al., 1993), serum response factor (Rivera et al., 1993) and Nur77 (Fisher and Blenis, 1996) in vitro, but direct evidence for the involvement of p90rsk1 in the phosphorylation of these transcription factors in vivo has not been provided yet.
In the present study, we report that p90rsk1 is a kinase that phosphorylates IκBα both in vitro and in vivo. p90rsk1 phosphorylates IκBα at Ser32, an event that has been shown to be pivotal for the induction of polyubiquitination and degradation of IκBα and subsequent activation of NFκB. Moreover, we show that p90rsk1 physically associates with IκBα in vivo. Ectopic expression of a dominant‐negative mutant of p90rsk1 in cells that are normally responsive to TPA, i.e. adenovirus type 5 early region 1 (Ad5E1)‐transformed baby rat kidney (BRK) cells, inhibits degradation of IκBα after TPA stimulation. Overexpression of dominant‐negative p90rsk1 does not interfere with IκBα degradation in response to TNF‐α. Consistently, in cells that have low levels of p90rsk1, i.e. adenovirus type 12 early region 1 (Ad12E1)‐transformed BRK cells, NFκB cannot be activated by TPA. These data prove that p90rsk1 is an essential kinase required for TPA‐induced phosphorylation and subsequent degradation of IκBα in vivo. Furthermore, these data suggest that multiple independent NFκB activation pathways exist, each requiring distinct IκBα kinases.
NFκB is not activated in response to TPA in Ad12E1‐transformed BRK cells
Identification of kinases involved in the phosphorylation of IκBα may come from studying cell types in which the regulation of NFκB is disturbed. Previously, we have shown that Ad12E1‐transformed BRK cells have reduced levels of NFκB and hardly any KBF1 (the homodimer of p50‐NFκB1) due to inefficient processing of p105‐NFκB1 (Schouten et al., 1995). The reduced processing of p105‐NFκB1 can be caused either by Ad12E1A‐mediated interference with the initiation of p105‐NFκB1 processing or by inhibition of the activity of the 26S proteasomes, in which p105‐NFκB1 processing takes place. The processing of p105‐NFκB1 and ligand‐induced degradation of IκBα are thought to be coupled processes, both occurring in the 26S proteasome (reviewed in Thanos and Maniatis, 1995). Therefore, we used Ad12E1‐transformed BRK cells to study their responsiveness to NFκB‐inducing agents and compared them with Ad5E1‐transformed BRK cells, which have high levels of NFκB. Ad5E1‐ and Ad12E1‐transformed BRK cells were treated with TNF‐α or TPA and tested for the induction of the DNA‐binding activity of NFκB to the H2TF1 element of the mouse MHC class I promoter in an electrophoretic mobility shift assay (EMSA). Figure 1A shows that treatment with TNF‐α caused rapid induction of NFκB, both in Ad5E1‐ (lanes 1–5) and Ad12E1‐transformed (lanes 6–10) BRK cells. The kinetics of NFκB induction in each cell type was not significantly different, although the absolute level of DNA‐bound NFκB was significantly lower in Ad12E1‐transformed cells as compared with Ad5E1‐ or non‐transformed BRK cells. This is in agreement with the finding that Ad12E1‐transformed cells contain reduced amounts of NFκB due to inefficient p105‐NFκB1 processing (Schouten et al., 1995). In contrast, TPA treatment activated NFκB in Ad5E1‐transformed BRK cells (lanes 11–15), but not in Ad12E1‐transformed BRK cells (lanes 16–20). In this particular experiment, the cells were treated for up to 1 h with TPA. Treatment of Ad12E1‐transformed cells for longer periods of time (up to 4 h) also did not cause NFκB activation (data not shown). This excludes the possibility that TPA‐induced activation of NFκB is delayed in Ad12E1‐transformed cells.
Next, we examined whether the observed TNF‐α‐ and TPA‐induced activation of NFκB was caused by enhanced degradation of its inhibitor IκBα. Therefore, AdE1‐transformed cells were treated with these agents for various time periods and extracts from the treated cells were assayed on a Western blot. Figure 1B shows that TNF‐α treatment of both Ad5E1‐ (lanes 1–5) and Ad12E1‐transformed BRK cells (lanes 6–10) caused rapid degradation of IκBα. In the course of the experiment, newly synthesized IκBα appeared (lanes 5 and 10). Since the expression of IκBα is controlled by NFκB (Arenzana‐Seisdedos et al., 1995, and references therein), this experiment showed that TNF‐α‐induced NFκB is transcriptionally active in both Ad5E1‐ and Ad12E1‐transformed cells. In contrast, TPA treatment only induced degradation of IκBα in Ad5E1‐transformed cells (lanes 11–15), whereas in Ad12E1‐transformed cells (lanes 16–20) the levels of IκBα remained unaffected. These data show that the TPA‐induced degradation of IκBα and subsequent activation of NFκB is inhibited in Ad12E1‐transformed cells. This is not due to inactivation of the 26S proteasome, since the induction of IκBα degradation and NFκB activation by TNF‐α was not abolished in these cells. Therefore, it is likely that earlier events in the pathway, e.g. TPA‐induced phosphorylation of IκBα, are affected in Ad12E1‐transformed cells.
Ad12E1‐transformed cells have reduced levels of p90rsk1
Ligand‐induced degradation of IκBα is triggered by its phosphorylation. Since it is the TPA‐induced degradation of IκBα that is blocked in Ad12E1‐transformed cells, we screened AdE1‐transformed cells for kinases that are responsive to TPA. The Western blot of Figure 2A (upper panel) shows that the levels of mitogen‐inducible p90rsk1 are much lower in Ad12E1‐transformed BRK cells (lanes 3 and 4) than in Ad5E1‐transformed BRK cells (lanes 1 and 2). The antibody recognized three distinct forms of p90rsk1, which represent differentially phosphorylated forms of p90rsk1 (see below). The transformed cells were also tested for the presence of the mitogen‐inducible ‘extracellularly regulated kinases’ p44erk1 and p42erk2. Neither p44erk1 nor p42erk2 protein levels were affected in AdE1‐transformed BRK cells (Figure 2A, lower panel). Furthermore, the p90rsk isoforms p90rsk2 and p90rsk3 (Moller et al., 1994; Zhao et al., 1995; Xing et al., 1996) were not down‐regulated in Ad12E1‐transformed BRK cells (Figure 2B). This shows that specifically the amount of p90rsk1 was reduced in Ad12E1‐transformed cells. Down‐regulation of p90rsk1 in Ad12E1‐transformed cells occurred at the level of RNA (data not shown).
The correlation between low levels of p90rsk1 and lack of TPA inducibility of NFκB in Ad12E1‐transformed cells suggests that p90rsk1 might be required for this process. However, based on these data, we cannot exclude that Ad12E1‐transformed cells are not responsive to TPA at all, for example due to lack of expression of certain protein kinase C isoforms. Therefore, we treated AdE1‐transformed BRK cells with TPA and determined the phosphorylation state of p90rsk1 by measuring the formation of hyperphosphorylated p‐p90rsk1 on a Western blot. Inactive hypophosphorylated p90rsk1 has a higher mobility in SDS–PAGE than TPA‐activated hyperphosphorylated p‐p90rsk1 (Grove et al., 1993). Figure 2C shows that TPA treatment induced the formation of hyperphosphorylated p‐p90rsk1 in Ad5E1‐transformed cells (lanes 1–5). Moreover, TPA treatment caused hyperphosphorylation of the low level of p90rsk1 in Ad12E1‐transformed cells (lanes 6–10). All three distinct forms of p90rsk1 were converted into one discrete form of p‐p90rsk1, suggesting that the three forms of p90rsk1 in unstimulated cells represent differentially phosphorylated forms of p90rsk1. In contrast, treatment of Ad5E1‐transformed cells with TNF‐α did not cause formation of p‐p90rsk1 (Figure 2D, lanes 1–5). Apparently, p90rsk1 is not part of the signal transduction route that is activated by TNF‐α in AdE1‐transformed cells. These data demonstrate that both Ad5E1‐ and Ad12E1‐transformed cells can be responsive to TPA.
p90rsk1 phosphorylates IκBα in vivo at Ser32
To investigate whether p90rsk1 is involved directly in the TPA‐induced phosphorylation of IκBα, we transiently transfected COS1 cells with a vector encoding haemagglutinin (HA)‐tagged p90rsk1 or, as a control, with a vector encoding an HA‐tagged kinase that acts directly upstream of p90rsk1, i.e. p44erk1. Both kinases were activated by treating the transfected cells with TPA, and the kinases were purified by immunoprecipitation. Subsequently, the activity of these kinases towards human GST–IκBα was determined in vitro. Figure 3 shows that p90rsk1 phosphorylated GST–IκBα (lane 2) whereas p44erk1 failed to do so (lane 3) under identical reaction conditions. Precipitation of an endogenous IκBα kinase with the 12CA5 α‐HA antibody was excluded, since no IκBα kinase activity was precipitated from mock‐transfected cells (lane 1). Neither p90rsk1 nor p44erk1 were found to phosphorylate GST alone (data not shown). As a control for the activity of the kinases, the kinase activity towards known substrates was assayed. p90rsk1 phosphorylated the ribosomal protein S6 (lane 5) and p44erk1 phosphorylated c‐Jun (lane 6), showing that both kinases were active. These data clearly show that p90rsk1 phosphorylates IκBα in vitro.
We next examined whether p90rsk1 is able to phosphorylate IκBα in vivo. COS1 cells were transiently transfected with a vector encoding wild‐type human IκBα with or without a vector encoding HA‐tagged p90rsk1 or HA‐tagged p44erk1, or an empty vector. Subsequently the cells were treated with TPA for either 10 or 20 min and assayed for phosphorylation of IκBα, as measured by a decreased mobility of IκBα in SDS–PAGE. Since the transfected IκBα is strongly overexpressed due to the presence of an SV40 origin of replication, IκBα phosphorylation by endogenous kinases was circumvented. As shown in Figure 4A, IκBα was not phosphorylated when an empty vector (lanes 1–3) or a p44erk1‐encoding vector (lanes 4–6) was co‐transfected, showing that phosphorylation of IκBα by putative endogenous IκBα kinases is limited under these assay conditions. In contrast, co‐expression of p90rsk1 did result in phosphorylation of IκBα (lanes 7–9), giving rise to the slower migrating form of IκBα, p‐IκBα. Phosphorylation of IκBα occurred even in untreated cells (lane 7), most likely due to the high levels of p90rsk1 in the transiently transfected cells. IκBα phosphorylation was enhanced further by TPA treatment (lanes 8 and 9). Both HA‐tagged p44erk1 and HA‐tagged p90rsk1 were highly expressed (data not shown).
In the experiments discussed above, both IκBα and p90rsk1 were strongly overexpressed. In order to test whether p90rsk1 can phosphorylate IκBα under more physiological circumstances, we transfected Ad12E1‐transformed cells (which express low levels of p90rsk1) with human IκBα and HA‐tagged p90rsk1. The transfected human IκBα and the endogenously expressed rat IκBα ran with different mobilities in SDS–PAGE, which allowed us to study the TPA‐induced phosphorylation of the two proteins separately. Figure 4B shows that neither the endogenous rat IκBα (rIκBα) nor the transfected human IκBα (hIκBα) was phosphorylated by endogenous IκBα kinases following TPA treatment (lanes 1–4). In contrast, co‐expression of p90rsk1 caused rapid phosphorylation of hIκBα in response to TPA (lanes 5–8). On the other hand, the phosphorylation of endogenous rIκBα was low in these transfected cells (lanes 5–8). The Ad12E1‐transformed cells had been transfected with an efficiency of <10%. Thus, <1 out of 10 cells co‐expressed significant amounts of p90rsk1 (transfected HA‐tagged p90rsk1) and endogenous rIκBα. In contrast, almost all transfected cells (<10% of the total culture) will co‐express hIκBα and HA‐tagged p90rsk1, when HA‐tagged p90rsk1 and hIκBα are co‐transfected (lanes 5–8). This explains why p90rsk1‐mediated phosphorylation of IκBα was observed only with hIκBα (lanes 5–8). It should be noted that co‐transfection of rIκBα and p90rsk1 into COS cells caused TPA‐induced phosphorylation of rIκBα (data not shown). Thus p90rsk1 can phosphorylate IκBα in vivo.
Stimulation of cells with NFκB‐inducing agents causes phosphorylation of IκBα at Ser32 and Ser36. Phosphorylation of these residues is a pivotal step in the activation of NFκB, since mutation of either one or both of these serines into alanines abolishes ligand‐induced degradation of IκBα (Brockman et al., 1995; Brown et al., 1995; Traenckner et al., 1995; Whiteside et al., 1995; DiDonato et al., 1996). In order to determine whether Ser32 and/or Ser36 was phosphorylated by p90rsk1, we tested the p90rsk1 kinase activity towards human wild‐type IκBα (wt‐IκBα) and its mutant forms 32A‐IκBα, 36A‐IκBα and 32A36A‐IκBα, in which Ser32, Ser36 or both were replaced by alanines (Whiteside et al., 1995). Figure 4C shows that p90rsk1 phosphorylated wt‐IκBα (lanes 1–3) in response to TPA, giving rise to p‐IκBα. Mutation of both Ser32 and Ser36 (lanes 10–12) or mutation of Ser32 alone (lanes 4–6) abolished p90rsk1‐mediated phosphorylation of IκBα. In contrast, mutation of Ser36 did not inhibit the formation of p‐IκBα (lanes 7–9). p90rsk1 phosphorylates serine residues in the consensus sequence RXXS, where X may be any amino acid (Erikson and Maller, 1988). In agreement with this, Ser32, but not Ser36, of IκBα lies in such a consensus sequence: RHDS. This specific sequence is conserved in IκBα of human, mouse, rat, pig and chicken.
IκBα and p90rsk1 associate in vivo
Next, we studied whether p90rsk1 physically associates with IκBα. IκBα was transiently expressed in COS1 cells with or without HA‐tagged p90rsk1 or HA‐tagged p44erk1. Prior to cell lysis, the transfected cells were treated with TPA for 15 min, or left untreated. IκBα and p90rsk1 were immunoprecipitated under mild conditions and the immunoprecipitates were assayed on a Western blot. Figure 5A shows that p90rsk1 co‐precipitated with IκBα (lane 3) and vice versa (lanes 7, 8 and 10). TPA treatment of the transfected cells caused dissociation of the IκBα–p90rsk1 complex (lanes 4, 9 and 11), indicating that this complex is disrupted upon p90rsk1‐mediated phosphorylation of IκBα. Co‐precipitation with IκBα was specific for p90rsk1, since p44erk1 did not co‐precipitate (Figure 5B, lanes 1 and 2). Cross‐reactivity of the antibodies was excluded, since the antibodies SC231 (α‐p90rsk1; Figure 5A, lane 5) and 12CA5 (α‐HA; Figure 5B, lane 2) did not precipitate IκBα in the absence of HA‐tagged p90rsk1. Furthermore, the α‐IκBα antibody SC371 did not cross‐react with HA‐tagged p90rsk1 (Figure 5A, lane 2). These data show that p90rsk1 and IκBα form a complex in vivo.
Stable overexpression of dominant‐negative p90rsk1 inhibits degradation of IκBα in response to TPA
To determine whether p90rsk1 is a physiological IκBα kinase, we have inactivated this kinase in Ad5E1 cells, by overexpressing a dominant‐negative mutant. p90rsk1 contains two distinct kinase domains. The amino‐terminal kinase domain is involved in substrate phosphorylation, whereas both domains contribute to autophosphorylation (Bjorbaek et al., 1995). With site‐directed mutagenesis, the ATP‐binding site of the amino‐terminal kinase domain was disrupted by replacing aspartic acid on position 205 in the DFG consensus sequence by asparagine (D205N) (van den Heuvel and Harlow, 1993). As expected, this mutant does not phosphorylate IκBα in vitro and in vivo (data not shown). Subsequently, Ad5E1‐transformed BRK cells were stably transfected with a vector encoding both HA‐tagged p90rsk1‐D205N and the neomycin resistance gene. As a control, the same type of cells were transfected with a vector encoding only the neomycin resistance gene. Stable cell lines were established and screened for expression of p90rsk1‐D205N. The expression levels were estimated to be 10‐ to 25‐fold higher than the endogenous p90rsk1 (Figure 6A). In cell lines expressing p90rsk1‐D205N, TPA‐induced degradation of IκBα was inhibited, whereas in the control cell lines BX‐D11 (Figure 6B, lanes 1–6) and BX‐D3 (data not shown) degradation was not affected. In contrast, degradation of IκBα in response to TNF‐α was not inhibited by the expression of p90rsk1‐D205N. These experiments show that p90rsk1 is an essential kinase that is required for TPA‐inducible phosphorylation and subsequent degradation of IκBα in vivo. Moreover, these data suggest that different NFκB‐inducing signals activate distinct IκBα kinases.
p90rsk1 associates with p105‐NFκB1 and phosphorylates its C‐terminus
Treatment of cells with NFκB‐activating agents not only causes phosphorylation and degradation of IκBα, but also leads to phosphorylation and processing of p105‐NFκB1 (Mellits et al., 1993; Mercurio et al., 1993; Naumann and Scheidereit, 1994; MacKichan et al., 1996). Therefore, we examined whether p90rsk1 is also involved in the phosphorylation of p105‐NFκB1. The human p105‐NFκB1 protein contains two putative p90rsk1 phosphorylation sites RXXS, i.e. Ser338 and Ser938. Ser338 is located in the RHD of the protein and Ser938 in the C‐terminal part of the protein. In order to discriminate between these two serine residues, we tested two forms of p105‐NFκB1 as substrates in a p90rsk1 kinase assay, i.e. full‐length Myc‐tagged p105‐NFκB1 and Myc‐tagged p97, a C‐terminal deletion mutant of p105‐NFκB1 lacking Ser938 (Fan and Maniatis, 1991). COS1 cells were transfected with a vector expressing HA‐tagged p90rsk1 or HA‐tagged p44erk1, and treated with TPA for 15 min. The kinases were purified by immunoprecipitation and their kinase activities were determined towards the different substrates. Figure 7A shows that full‐length p105 was phosphorylated by p90rsk1 (lane 2). Most likely, phosphorylation of p105‐NFκB1 occurred at Ser938 and not at Ser338, since p97 was not phosphorylated by p90rsk1 (lane 5). p44erk1 failed to phosphorylate either p105 (lane 3) or p97 (lane 6). The 12CA5 α‐HA antibody did not cross‐react with an endogenous p105‐NFκB1 kinase, since no kinase activity was precipitated from the mock‐treated cells (lanes 1 and 4). p90rsk1 and p44erk1 phosphorylated the S6 protein of the ribosomal 40S subunit (lane 7) and c‐Jun (lane 8), respectively, indicating that both kinases were active.
As p90rsk1 specifically phosphorylates the C‐terminus of p105‐NFκB1, we also determined whether p90rsk1 binds to p105‐NFκB1. COS1 cells were transfected with pCMV‐105T or pCMV‐97T (encoding Myc‐tagged p105‐NFκB1 or Myc‐tagged p97‐NFκB1) with or without a vector expressing HA‐tagged p90rsk1 or HA‐tagged p44erk1. Co‐immunoprecipitations were performed under mild conditions and the immunoprecipitates were analysed on a Western blot. Figure 7B shows that p105 and p97 co‐precipitated with p90rsk1 (lanes 14 and 16) and vice versa (lanes 6 and 9). Only the inactive hypophosphorylated form of p90rsk1 associated with p105 (lane 9). Furthermore, TPA treatment of the transfected cells caused dissociation of the p90rsk1–p105 complex (lanes 10 and 17). On the other hand, the p90rsk1–p97 complex did not dissociate following TPA treatment (lanes 7 and 15). Instead, a hyperphosphorylated form of p90rsk1 (p‐p90rsk1) co‐precipitated with p97 after TPA treatment (lane 7). This suggests that TPA‐induced disruption of the p105–p90rsk1 complex is caused by p90rsk1‐mediated phosphorylation of the C‐terminus of p105‐NFκB1 (the part of p105 that is missing in p97), rather than by phosphorylation of p90rsk1 itself. Thus, the p90rsk1–p97 complex is more stable than the p90rsk1–p105 complex, explaining why p90rsk1 co‐precipitated less efficiently with p105 than with p97 (compare lanes 9 and 16 with lanes 6 and 14). Figure 7C shows that p97 specifically associates with p90rsk1 (lanes 4 and 9), whereas no complex formation can be detected with p44erk1 (lanes 5 and 10). Cross‐reaction of the antibodies was excluded, since p90rsk1 could not be precipitated with the 9E10 α‐Myc antibody in the absence of Myc‐tagged p97 (lane 1) and p97 could not be precipitated with the 12CA5 α‐HA antibody in the absence of HA‐tagged p90rsk1 (lane 8). More importantly, co‐immunoprecipitation of p90rsk1 with p97 was completely abolished by addition of 0.1% SDS (Figure 7B, lane 8). Thus, p90rsk1 physically associates with p105‐NFκB in vivo.
In the present study, we provide evidence that the mitogen‐inducible kinase p90rsk1 is an IκBα kinase. It phosphorylates IκBα at Ser32 in vivo. Moreover, p90rsk1 physically associates with IκBα. Interestingly, the p90rsk1–IκBα complex dissociates upon treatment of cells with TPA. Based on the finding that TPA‐induced dissociation of the p90rsk1–p105‐NFκB1 complex occurred only when the p90rsk1 phosphorylation site in the C‐terminus of p105‐NFκB1 was present (see below), we suggest that dissociation of the p90rsk1–IκBα complex is caused by p90rsk1‐induced phosphorylation of IκBα. Impairing the function of p90rsk1 in vivo, by expression of a dominant‐negative mutant, inhibits the degradation of IκBα in response to TPA. Concomitant with the role of p90rsk1 as an IκBα kinase required for NFκB activation in response to mitogenic agents, NFκB cannot be activated by TPA in cells that have low levels of p90rsk1.
Phosphorylation of Ser32 has been reported previously to be pivotal for the ubiquitination of IκBα and subsequent degradation of IκBα in response to NFκB‐activating agents (Brockman et al., 1995; Brown et al., 1995; Chen et al., 1995; Traenckner et al., 1995; Whiteside et al., 1995; DiDonato et al., 1996). However, p90rsk1‐mediated phosphorylation of IκBα does not cause its degradation. This can be explained by the fact that IκBα has to be phosphorylated not only at Ser32 but also at Ser36 before degradation occurs (Brockman et al., 1995; Brown et al., 1995; Traenckner et al., 1995; Whiteside et al., 1995; DiDonato et al., 1996). Since p90rsk1 failed to phosphorylate Ser36, the p90rsk1‐phosphorylated IκBα was not degraded. This also explains why co‐transfection of a vector expressing p90rsk1 and an NFκB‐dependent luciferase reporter does not lead to increased luciferase activity (data not shown). These data thus suggest that multiple IκBα kinases are required for the induction of IκBα degradation. This assumption is attractive, since it allows fine‐tuning of the regulation of IκBα degradation and NFκB activation. Furthermore, it explains why IκBα is phosphorylated in response to a wide variety of stimuli that activate different subsets of kinases. Indeed dominant‐negative p90rsk1 interferes with TPA‐induced, but not with TNF‐α‐induced degradation of IκBα in vivo. This underlines the existence of multiple activation pathways, that independently mediate IκBα phosphorylation. Fine‐tuning of NFκB activation is enhanced further by the existence of multiple IκB proteins, e.g. IκBα and IκBβ. In contrast to IκBα, the degradation of IκBβ is induced only by a limited number of stimuli, i.e. lipopolysaccharide and interleukin‐1, whereas others, such as TPA, have no effect (Thompson et al., 1995). IκBβ degradation is induced by phosphorylation of serine residues 19 and 23 (DiDonato et al., 1996). Interestingly, these serine residues do not lie in the p90rsk1 consensus sequence RXXS, which might explain why IκBβ is not degraded in response to TPA.
Thus far, several kinases have been reported to be involved in ligand‐induced phosphorylation of IκBα and activation of NFκB. The double‐stranded (ds) RNA‐activated p86 protein kinase (PKR) phosphorylates IκBα in response to dsRNA in vivo (Kumar et al., 1994; Maran et al., 1994). Antisense PKR abolishes dsRNA‐induced degradation of IκBα, but it does not affect IκBα degradation induced by other stimuli. This rules out the requirement of PKR for NFκB activation in general (Maran et al., 1994). Furthermore, p70s6k has been reported to be involved in the activation of NFκB (Lai and Tan, 1994), although this kinase has not been shown to phosphorylate IκBα directly in vivo. More recently, it has been reported that IκBα is phosphorylated by casein kinase II (CKII), both in vitro and in vivo (Barroga et al., 1995; Lin et al., 1996; McElhinny et al., 1996; Schwarz et al., 1996). Ser32 and Ser36 lie within the CKII consensus sequence S/TXXZ, in which Z must be an acidic amino acid or a phosphorylated serine or threonine (Pinna, 1990). The X residues preferentially should be acidic amino acids, but for Ser32 and Ser36 the residues represented by X are not acidic, suggesting that the consensus sequence is imperfect. In agreement with this, CKII‐mediated phosphorylation of IκBα was found only at serine and threonine residues located in the C‐terminus of IκBα, whereas no phosphorylation occurred within the first 60 amino acids. Furthermore, CKII phosphorylated IκBα constitutively (Barroga et al., 1995; Lin et al., 1996; McElhinny et al., 1996; Schwarz et al., 1996). Finally, a large multisubunit ubiquitination‐dependent protein kinase that phosphorylates IκBα has been isolated from HeLa cells. Interestingly, this kinase complex phosphorylated IκBα at Ser32 and/or Ser36 (Chen et al., 1996), the same residues that are phosphorylated upon stimulation of cells with NFκB‐inducing agents. However, the identity of the kinase(s) is not known yet. Recently, an alternative mechanism to inactivate IκBα was reported. Stimulation of Jurkat T cells with pervanadate, or reoxygenation of hypoxic HeLa cells, led to inactivation of IκBα due to its phosphorylation at Tyr42, although this inactivation was not linked to IκBα degradation (Imbert et al., 1996).
Several lines of evidence indicate that the Ras/MAP kinase cascade mediates ligand‐induced activation of NFκB. First, several protein kinase C (PKC) isoenzymes induce phosphorylation of IκBα (Diaz‐Meco et al., 1993, 1994; Steffan et al., 1995; Janosch et al., 1996; Lindholm et al., 1996), but PKC failed to phosphorylate IκBα directly (Janosch et al., 1996). Secondly, Ha‐Ras and Raf‐1 were shown to be required for ligand‐induced NFκB activation (Devary et al., 1993; Finco and Baldwin, 1993; Bertrand et al., 1995; Folgueira et al., 1996; Janosch et al., 1996). Contradicting results have been reported on the requirement for Raf‐1. The Raf‐1 kinase domain was shown to phosphorylate IκBα in vitro and to associate with IκBα in a yeast two‐hybrid system (Li and Sedivy, 1993). However, full‐length Raf‐1 failed to associate with IκBα or to phosphorylate it (Diaz‐Meco et al., 1994; Janosch et al., 1996). Based on the data presented in this study, we propose that the involvement of the Ras/MAP kinase cascade is transduction of mitogen‐induced signals to the IκBα kinase p90rsk1, which subsequently phosphorylates IκBα. Finally, NFκB is induced during the G0–G1 transition of the cell cycle (Baldwin et al., 1991). This coincides with the activation of p90rsk1 (Chen et al., 1991, 1992). Based on the data presented in this study, it is likely that p90rsk1 functions as an IκBα kinase in this particular process.
Concomitant with IκBα degradation, stimulation of cells with NFκB‐activating agents causes enhanced processing of p105‐NFκB1 (Mellits et al., 1993; Mercurio et al., 1993; Naumann and Scheidereit, 1994). This event is also preceded by phosphorylation of p105‐NFκB1 (Mellits et al., 1993; Naumann and Scheidereit, 1994). Interestingly, p90rsk1 phosphorylates p105‐NFκB1 and physically associates with it. Since both IκBα and p105‐NFκB1 contain an ARD, it is likely that this domain is responsible for association of these proteins with p90rsk1. Accordingly, p90rsk1 does not associate with p50‐NFκB1, the processed form of p105‐NFκB1 that lacks the ARD (data not shown). TPA‐induced activation of p90rsk1 causes dissociation of the p90rsk1–p105‐NFκB1 complex, which requires the presence of the C‐terminus of p105‐NFκB1, the part of p105‐NFκB1 that is phosphorylated by p90rsk1. This suggests that phosphorylation of the C‐terminus of p105 induces a change in its conformation, leading to dissociation of the p105–p90rsk1 complex. The assumed p90rsk1‐induced conformational change in p105‐NFκB1 might also commit its C‐terminus to ubiquitination and degradation. Interestingly, Ad12E1‐transformed cells that express low amounts of p90rsk1 process p105‐NFκB inefficiently (Schouten et al., 1995), which suggests that p90rsk1 indeed is required for the initiation of p105‐NFκB1 processing.
Materials and methods
Cell culture and DNA transfection
Monolayers of COS1 cells were cultured in Dulbecco‘s modified Eagle's medium (DMEM), supplemented with 10% fetal calf serum and antibiotics. The BRK cell lines transformed by Ad5E1 and Ad12E1 are described elsewhere (Schouten et al., 1995). Monoclonal cell lines BX‐RSK‐M13 and BX‐RSK‐M15 were established by transfection of the Ad5E1‐transformed BRK cell line BXC22 with pcDNA3‐p90Rsk1‐D205N and subsequent selection for G418 resistance. Control monoclonal cell lines BX‐D3 and BX‐D11 were established by transfection of BXC22 with parental vector pcDNA3 and subsequent selection for G418 resistance. Monolayers of AdE1‐transformed cells were cultured in Eagle's minimal essential medium (MEM) supplemented with 10% newborn calf serum and antibiotics. Tissue culture media and sera were purchased from Gibco (Grand Island, NY). All tissue culture plastics were obtained from Greiner (Nürtingen, Germany). Two hours prior to transfection, cells were plated at 30% density on 90 mm or 30 mm dishes. Two hours after plating, cells were incubated for 15 h with calcium phosphate‐precipitated DNA as described previously (van der Eb and Graham, 1980). Subsequently, the transfected cells were cultured for 30 h, cells were harvested and extracts were prepared. Treatment with 100 ng/ml TPA (Sigma) was performed for the indicated periods of time.
Electrophoretic mobility shift assay
Preparation of whole cell extracts, binding reactions and electrophoresis were performed essentially as described elsewhere (Bernards, 1991). The double‐stranded H2TF1 oligonucleotide dCCCAGGGCTGGGGATTCCCCATTGCA was labelled by filling in recessed 3′ ends with [α–32P]dCTP and the Klenow fragment of DNA polymerase I. After electrophoresis, the gels were dried and exposed to Kodak XAR‐5 film at −80°C.
Plasmids, proteins and antibodies
The vectors expressing human wt‐IκBα and the mutant IκBα proteins 32A‐IκBα, 36A‐IκBα and 32A36A‐IκBα have been described elsewhere (Whiteside et al., 1995). The pCMV‐rIκBα expression vector was constructed as follows: the full‐length IκBα cDNA from rat (RL/IF1) was isolated from Bluescript‐IκBα (Tewari et al., 1992) by digestion with HindIII and EcoRI. Subsequently, the full‐length cDNA was ligated into a HindIII–EcoRI‐digested pcDNA3 vector (Invitrogen), giving rise to pCMV‐rIκBα. Expression vectors pCMV‐p105T and pCMV‐p97T (Fan and Maniatis, 1991), which encode Myc‐tagged p105‐NFκB1 and a C‐terminally deleted form of p105‐NFκB1, respectively, were kindly provided by Dr T.Maniatis. The plasmids pMT2‐Rsk1‐epi, encoding HA‐tagged p90rsk1 (Grove et al., 1993) and pcDNA‐p44‐tag, encoding HA‐tagged p44erk1 (Meloche et al., 1992), were kind gifts of Drs P.J.Coffer and J.Pouysségur, respectively. pcDNA3‐Rsk1‐epi was constructed by cloning a 3.3 kb EcoRI fragment of pMT2‐Rsk1‐epi into the EcoRI site of pcDNA3 (Invitrogen). Site‐directed in vitro mutagenesis of p90rsk1 was performed using the Altered Sites System, according to instructions of the manufacturer (Promega). A 2410 bp BamHI–SalI fragment of pMT2‐Rsk1‐epi was cloned into the multiple cloning site of the pSelect1 vector. The ATP‐binding site of the amino‐terminal kinase domain of p90rsk1 was mutated using oligonucleotide 5′‐GCCACATCAAACTCACAAATTTTGGCCTGAGCAAGG‐3′ (pSelect1‐Rsk1‐D205N). This mutation replaces aspartic acid on position 205 by asparagine and introduces a diagnostic ApoI restriction site. Full‐length HA‐Rsk1‐D205N was created by replacing a 725 bp BstEII–PflMI fragment from pcDNA3‐Rsk1‐epi by the BstEII–PflMI fragment from pSelect1‐Rsk1‐D205N. Purified GST–IκBα and c‐Jun proteins were purchased from Santa Cruz Biotechnology Inc. and Promega, respectively. The 40S ribosomes were isolated from rat liver and were a gift from Dr J.Dijk. The mouse monoclonal antibodies 12CA5 α‐HA and 9E10 α‐Myc are described elsewhere (Wilson et al., 1984; Evan et al., 1985). The antibodies SC231 α‐RSK1, SC1430 α‐RSK2, SC1431 α–RSK3 and SC371 α‐IκBα were purchased from Santa Cruz Biotechnology Inc.
Cells were washed twice with phosphate‐buffered saline (PBS), and lysed and scraped in mild buffer [50 mM Tris pH 7.4, 250 mM NaCl, 5 mM EDTA and 0.1% Triton X‐100, supplemented with 1 mM phenylmethylsulfonyl fluoride (PMSF), 1 mg/ml Na4P2O7 and 0.1 mg/ml trypsin inhibitor, NaF, β‐glycerophosphate and Na3VO4). After 30 min incubation on ice, the lysates were cleared by centrifugation. Immunoprecipitations were performed by incubation of extracts derived from half a 90 mm dish with 1 μg of specific antibody and 30 μl of 10% protein A beads for 2 h. After extensive washing, the immunoprecipitates were subjected to Western blotting.
Cells were washed twice with PBS, and lysed and scraped in RIPA (1% NP‐40, 0.5% sodium deoxycholate and 0.1% SDS in PBS, supplemented with 1 mM PMSF, 1 mg/ml Na4P2O7 and 0.1 mg/ml trypsin inhibitor, NaF, β‐glycerophosphate and Na3VO4). After 30 min incubation on ice, the lysates were cleared by centrifugation. Protein concentrations were determined by the Bio‐Rad protein assay. Whole cell extracts or immunoprecipitates were fractionated by SDS–PAGE on 10% gels. Proteins were transferred onto Immobilon‐P membranes and incubated with specific antibodies as indicated. The secondary antibodies were horseradish peroxidase‐coupled goat anti‐rabbit, goat anti‐mouse and donkey anti‐goat IgG (Santa Cruz Biotechnology Inc.). The antibody complexes were visualized with the ECL detection system according to the manufacturer's protocol (Amersham, UK).
Kinase assays were performed essentially as described previously (Chen and Blenis, 1990), with some minor modifications. TPA‐treated cells were lysed and scraped in lysis buffer (1% NP‐40, 10 mM KH2PO4, 1 mM EDTA, 5 mM EGTA, 10 mM MgCl2 and 50 mM β‐glycerophosphate pH 7.2, supplemented with 1 mM PMSF, 1 mg/ml Na4P2O7 and 0.1 mg/ml trypsin inhibitor, NaF and Na3VO4). After 30 min incubation on ice, the lysates were cleared by centrifugation. Subsequently, HA‐tagged p90rsk1 and HA‐tagged p44erk1 were purified by immunoprecipitation with 12CA5 α‐HA. Kinase reactions were performed with the indicated substrates (purified GST–IκBα, c‐Jun, 40S ribosomes, purified Myc‐tagged p105‐NFκB1 or purified Myc‐tagged p97‐NFκB1) in kinase buffer [20 mM HEPES pH 7.2 and 10 mM MgCl2, supplemented with 0.1 mg/ml trypsin inhibitor, 0.1 mg/ml Na3VO4, 50 μM ATP, 0.1 mg/ml bovine serum albumin (BSA) and 2 mM dithiothreitol (DTT)] in the presence of [γ‐32P]ATP for 30 min at 30°C. The phosphorylated proteins were fractionated by SDS–PAGE on 10% gels. The gels were dried and exposed to Kodak XAR‐5 film at −80°C.
We thank Dr T.Maniatis for kindly providing the human p97T and p105T expression vectors and Dr R.Taub for the gift of pBluescript‐IκBα. We also thank Drs P.J.Coffer and J.Pouysségur for the gift of plasmids pMT2‐Rsk1‐epi and pcDNA‐p44‐tag, respectively. We thank Dr J.Dijk for kindly providing us with the 40S ribosomes from rat liver. Ron Schouten is kindly acknowledged for photography. Finally, we thank Dr H.van Ormondt for critically reading the manuscript. This work was supported by a grant from the Dutch Cancer Society. S.T.W. is the recipient of a Wellcome Travelling Fellowship.
↵† G.J.Schouten and A.C.O.Vertegaal contributed equally to this work
- Copyright © 1997 European Molecular Biology Organization