Glycogen Synthase Kinase 3β Functions To Specify Gene-Specific, NF-κB-Dependent Transcription

Abstract

Loss of glycogen synthase kinase 3β (GSK-3β) in mice results in embryonic lethality via hepatocyte apoptosis. Consistent with this result, cells from these mice have diminished nuclear factor κB (NF-κB) activity, implying a functional role for GSK-3β in regulating NF-κB. Here, we have explored mechanisms by which GSK-3β may control NF-κB function. We show that cytokine-induced IκB kinase activity and subsequent phosphorylation of IκBα, p105, and p65 are not affected by the absence of GSK-3β activity. Furthermore, nuclear accumulation of p65 following tumor necrosis factor treatment is unaffected by the loss of GSK-3β. However, NF-κB DNA binding activity is reduced in GSK-3β null cells and in cells treated with a pharmacological inhibitor of GSK-3. Expression of certain NF-κB-regulated genes, such as IκBα and macrophage inflammatory protein 2, is minimally affected by the absence of GSK-3β. Conversely, we have identified a subset of NF-κB-regulated genes, including those for interleukin-6 and monocyte chemoattractant protein 1, that require GSK-3β for efficient expression. We show that efficient localization of p65 to the promoter regions of the interleukin-6 and monocyte chemoattractant protein 1 genes following tumor necrosis factor alpha treatment requires GSK-3β. Therefore, GSK-3β has profound effects on transcription in a gene-specific manner through a mechanism involving control of promoter-specific recruitment of NF-κB.

The transcription factor nuclear factor κB (NF-κB) is composed of a number of structurally related REL family proteins that form DNA-binding dimers and control transcription of certain target genes. The NF-κB subunits are RelA (also called p65), RelB, c-Rel, NF-κB1 (p105), and NF-κB2 (p100), with the last two processed to release p50 or p52, respectively (15, 17). NF-κB proteins can form a variety of homo- and heterodimers, all of which bind similar DNA elements. A primary mechanism for the inhibition of NF-κB is the interaction with members of the IκB family of proteins (15-17). Signal-dependent phosphorylation and degradation of IκB release NF-κB for accumulation in the nucleus and subsequent control of gene-specific transcription. The IκB kinase (IKK) complex is the primary mediator of IκB phosphorylation and is activated by various stimuli including cytokines, such as tumor necrosis factor alpha (TNF-α) (17). In this respect, the IKK complex is critical for cytokine-mediated activation of NF-κB. This system of inducible activation of NF-κB is further controlled by posttranslational modifications such as phosphorylation of NF-κB subunits as well as interaction with transcriptional coactivators (17, 41, 42). Therefore, multiple events in the regulation of NF-κB activity suggest an extremely complex and context-dependent function for this transcription factor.

The two highly homologous glycogen synthase kinase 3 (GSK-3) proteins GSK-3α and GSK-3β are critical factors in the proper regulation of a wide variety of signaling proteins and transcription factors, including cyclin D1, c-Jun, NF-ATc, and β-catenin, among others (3, 5, 8, 12). In Wnt signaling, GSK-3β is a critical component of the adenomatous polyposis coli-β-catenin destruction complex and is especially important in initiating phosphorylation-dependent degradation of β-catenin (1, 30). Constitutively active GSK-3β maintains low levels of β-catenin in resting cells and thereby suppresses transcriptional activation by β-catenin-T-cell-specific transcription factor/lymphoid enhancer factor complexes. Importantly, GSK-3α can compensate for the loss of GSK-3β with respect to mediating β-catenin phosphorylation (13, 19). High levels of β-catenin that occur upon the blockade of both GSK-3 proteins are reported to inhibit NF-κB activity at the level of DNA binding, although the mechanism by which this occurs has not been established (10, 11).

Surprisingly, genetic targeting of GSK-3β (19) results in a phenotype similar to that of mice lacking the p65 NF-κB subunit or IKKβ (4, 22, 23). Specifically, ablation of these genes results in lethality at approximately embryonic day E13.5 due to TNF-α-dependent hepatocyte apoptosis. The initial report indicated that there were no defects in degradation of IκBα or nuclear translocation of p65 in fibroblasts lacking GSK-3β but that NF-κB DNA binding in gel shift assays as well as luciferase reporter activity were diminished (19). Since β-catenin levels are similar to wild-type levels in GSK-3β null fibroblasts, these observations suggested the possibility of an effect on NF-κB by GSK-3β that is independent of β-catenin.

Subsequently, numerous reports have implicated GSK-3β in the control of various signaling pathways that activate NF-κB, including regulation of NF-κB1/p105 stability as well as IKK activity (9, 32, 35). Additional work proposed that the p65 subunit of NF-κB is a direct target of GSK-3β kinase activity and that an NF-κB-dependent gene reporter response is consequently affected by these modifications (6, 33). The function of GSK-3β in signaling mechanisms that activate NF-κB as well as the resulting effects on NF-κB-mediated gene expression remain, however, unclear. This is further complicated by reports that fail to acknowledge the distinction between β-catenin-mediated effects on NF-κB and direct regulation of this transcription factor via the two GSK-3 isoforms.

Here, we have addressed the role of GSK-3β in control of NF-κB activation and target gene expression in mouse embryonic fibroblasts lacking GSK-3β as well as in nontransformed rat intestinal epithelial cells using a pharmacological inhibitor of GSK-3 activity. Neither of these cell lines exhibit elevated levels of β-catenin as a result of decreased GSK-3 activity compared to control cells and therefore provide diverse model systems in which to investigate the influence of GSK-3 on NF-κB. Importantly, we found no changes in TNF-α-induced IKK activation, subsequent degradation of IκBα, and nuclear localization of p65 in association with GSK-3β activity loss in either cell line. However, NF-κB DNA binding and luciferase reporter assay activity were reduced in the GSK-3β null cells, consistent with the report that GSK-3β is able to affect NF-κB function downstream of release from IκBα (19). Furthermore, we show that loss of GSK-3β specifically affects a subset of NF-κB-regulated genes during cytokine stimulation. We demonstrate by chromatin immunoprecipitation assays that GSK-3β is necessary for efficient localization of p65 to the promoter regions of these genes. These data indicate that GSK-3β is not a regulator of TNF-α-induced signaling to IKK in fibroblasts or intestinal epithelial cells. Furthermore, GSK-3β functions in a gene-specific manner to either facilitate access to promoter regions or allow direct promoter-specific recruitment of NF-κB-containing complexes.

MATERIALS AND METHODS

Cell culture and reagents.

Wild-type and GSK-3β null mouse embryonic fibroblasts (MEFs) (gift of J. Woodgett) were cultured in Dulbecco's modified Eagle's medium-H supplemented with 10% fetal calf serum, 50 U/ml penicillin, 50 μg/ml streptomycin, and 10 μg/ml neomycin. IEC-18 cells were grown in Dulbecco's modified Eagle's medium-H with additional 5% fetal calf serum, 50 U/ml penicillin, 50 μg/ml streptomycin, and 10 μg/ml neomycin and 0.1 U/ml insulin. All cell culture reagents were obtained from Sigma-Aldrich (St. Louis, MO). TNF-α was obtained from Promega (Madison, WI). Antibodies from Santa Cruz Biotechnology (Santa Cruz, CA) include β-tubulin (SC-9104), β-catenin (SC-7199), IκBα (SC-371), and p105 (SC-7178). Phospho-IκBα serine 32/36, phospho-p105 serine 927 and serine 933, phospho-p65 serine 536, and phospho-β-catenin ser33/37/thr41 were all from Cell Signaling Technology (Beverly, MA). IKKβ and TFIIB monoclonal antibodies were from Upstate Biotechnology (Waltham, MA) and Transduction Laboratories (Lexington, KY), respectively. SB216763 was obtained from Sigma-Aldrich.

Western analysis and kinase assays.

Cells were lysed in modified radioimmunoprecipitation buffer containing protease and phosphatase inhibitors (Sigma-Aldrich) for 10 min on ice and clarified by centrifugation at 14,000 × g for 10 min. Supernatants were quantitated by the Bradford method (Bio-Rad; Hercules, CA) and 20 μg of protein was denatured in sodium dodecyl sulfate loading buffer and fractionated on 4% to 12% NuPage acrylamide gels (Invitrogen; Carlsbad, CA). Following transfer onto nitrocellulose membrane and Ponceau S staining to confirm even protein loading, blots were incubated with appropriate antibodies in 5% milk/1% bovine serum albumin overnight at 4°C. Chemiluminescent detection was used to visualize target proteins (Amersham, Piscataway, NJ).

IKK activity assays were performed by first lysing cells in m-PER reagent (Pierce, Rockford, IL), centrifuging at 14,000 × g for 10 min and quantitating by Bradford assay. Cell extracts (250 μg) were then precleared with protein A/G Plus-agarose (Santa Cruz) for 2 h and incubated overnight at 4°C with IKKβ monoclonal antibody (Upstate Biotechnology). Immunoprecipitation with protein A/G Plus-agarose was followed with incubation with 2.0 μg glutathione S-transferase (GST)-IκBα (residues 1 to 54) fusion protein substrate in kinase buffer (25 mM Tris, pH 7.5, 5 mM β-glycerophosphate, 2 mM dithiothreitol, 0.1 mM sodium vanadate, 10 mM MgCl₂) in the presence of 200 uM ATP. Reaction mixtures were incubated at 30°C for 30 min, heat inactivated, and separated by polyacrylamide gel electrophoresis (PAGE). Coomassie staining of the gel was performed to confirm even GST-IκBα loading. Western analysis with IκBα-phospho-ser32/36 antibody provided an indication of kinase activity and IKKβ immunoblotting was performed to confirm efficient immunoprecipitation.

Electrophoretic mobility shift assays.

Cells were treated with appropriate reagents (i.e., 10 ng/ml TNF-α and/or 30 μM SB216763), nuclear and cytoplasmic extracts were obtained, and gel shift assays were performed as previously described (27). Briefly, an oligonucleotide corresponding to an NF-κB site in the H-2K^b gene was radiolabeled using [α-³²P]dCTP (Perkin Elmer). The probe was incubated with 5.0 μg of nuclear extract and 0.1 μg/μl poly(dI-dC) in DNA binding buffer (50 mM NaCl, 10 mM Tris, pH 7.6, 10% glycerol, 1 mM dithiothreitol, 0.5 mM EDTA) for 15 min at room temperature. For antibody supershift analysis, extracts were incubated 15 min at room temperature with 1 μg of antiserum before the addition of the radiolabeled gel shift probe. Reactions were separated using nondenaturing PAGE and visualized by autoradiography.

DNA affinity pull-down assays.

The DNA affinity pull-down assay was performed by incubating nuclear proteins (30 μg) in electrophoretic mobility shift assay DNA binding buffer containing 0.1 μg/ml poly(dI-dC) with 100 μl streptavidin MagneSphere paramagnetic beads at 4°C for 1 hour (Promega). Precleared extracts were then incubated for 1 hour at 4°C with 500 nM biotinylated DNA duplex (Integrated DNA Technologies, Coralville, IA) containing a NF-κB DNA binding sequence (H-2K^b site). DNA was then purified using a magnetic column and washed with DNA binding buffer four times. Purified protein complexes were then eluted into gel loading buffer, placed on a 4 to 12% NuPage gel, and immunoblotted with a p65-specific antibody (Rockland, Gilbertsville, PA).

ELISA-based DNA binding analysis.

As per the manufacturer's protocol (TransAm NF-κB; Active Motif, Carlsbad, CA), the DNA binding capacity of p65-containing NF-κB dimers was assessed using enzyme-linked immunosorbent assay (ELISA) plates containing fixed κB binding site consensus sequences. Briefly, 10 μg of nuclear extract was diluted into binding buffer and incubated for 1 h at room temperature. Following three washes, primary antibody specific for p65 was added to each well (1:1,000 dilution) and incubated again at room temperature for 1 hour. Horseradish peroxidase-conjugated secondary antibody incubation followed by addition of chromogenic substrate and measurement of optical density (450 nm; VersaMax Microplate Reader; Molecular Devices, Sunnyvale, CA) provided a quantitation of DNA bound p65. Samples from all time points were assessed in triplicate. Addition of oligonucleotides containing mutant or wild-type κB sites either had no effect or completely blocked p65 binding, respectively (K. A. Steinbrecher and A. S. Baldwin, unpublished data).

Real-time RT-PCR and transient transfection.

Trizol was used to extract RNA from appropriately treated cells and reverse transcription using oligo(dT) primers was performed using Superscript II as per the manufacturer's protocol (Invitrogen). Prior to reverse transcription, contaminating DNA was removed from RNA via treatment with Turbo DNase (Ambion, Austin, TX). QuantiTect SYBR Green reagents (QIAGEN) were used to determine the relative amount of macrophage inflammatory protein 2 (MIP-2) and monocyte chemoattractant protein 1 (MCP-1) mRNA (IDT, Coralville, IA; sequences available upon request) and Applied Biosystems Taqman Assay-On-Demand primer-probe sets were used for IκBα, interleukin-6 (IL-6), and glyceraldehyde-3-phosphate dehydrogenase. Reactions were performed on an ABI 7000 sequence detection system and relative quantification was determined using the ΔΔC_t method.

Wild-type (GSK-3β WT) and kinase-dead (GSK-3β KD) GSK-3β expression vectors were transiently transfected prior to extraction of RNA in some experiments. Briefly, GSK-3β-specific primers (5′-GCA TTT ATC ATT AAC CTA GCA CCC-5′ and 5′-ATT TTC TTT CCA AAC GTG ACC-3′) were used to amplify GSK-3β cDNA via RT-PCR with wild-type MEF RNA as the template. Primers specifying a point mutant in the catalytic domain (K85R) were then used to generate kinase-dead GSK-3β using site-directed mutagenesis (QuikChange; Stratagene, La Jolla, CA) (14). Transfections utilized Fugene 6 (Roche) according to the manufacturer's protocol.

ChIP.

Chromatin immunoprecipitation (ChIP) assays were performed using chromatin immunoprecipitation reagents according to the manufacturer's protocol (Upstate Biotechnology) with minor modifications. Mouse embryonic fibroblasts were treated with TNF-α, cross-linked with formaldehyde and washed extensively with phosphate-buffered saline. Cells were lysed and sonicated such that DNA fragment were 500 to 1,000 base pairs in length. Overnight incubation with p65 antibodies (Rockland) was followed by precipitation with salmon sperm DNA/protein A-agarose beads and washing. Immunoprecipitated DNA was decross-linked overnight in high-salt solution at 65°C and purified by spin column (QIAGEN).

Real-time PCR analysis was performed by normalizing to input DNA for each sample. PCRs were performed using the following primers which span κB sites at their promoters: IκBα forward, 5′TGG CGA GGT CTG ACT GTT GTG G3′, and reverse, 5′GCT CAT CAA AAA GTT CCC TGT GC3′; MIP-2 forward, 5′CAG GGC AGT AGA ATG AGG CAG G3′, and reverse, 5′AGG CTG AAG TGT GGC TGG AGT C 3′; MCP-1 forward, 5′CAC CCC ATT ACA TCT CTT CCC C, and reverse, 5′TGT TTC CCT CTC ACT TCA CTC TGT C3′; and IL-6 forward, 5′AAG CAC ACT TTC CCC TTC C3′, and reverse, 5′CTA TCG TTC TTG GTG GGC TC3′.

RESULTS

GSK-3β is not required for TNF-α-induced IKK activation and IκBα degradation.

Given the different conclusions that have been generated regarding the effect of loss of GSK-3β on NF-κB activation (i.e., no effect on IκBα degradation versus complete inhibition of IKK activity), we first addressed the effect of GSK-3β ablation on IKK activity in GSK-3β null MEFs. IKKβ was immunoprecipitated from extracts of unstimulated cells and from cells treated with TNF for 15 and 30 min. Kinase activity of the isolated IKK complex towards a GST-IκBα fusion protein substrate was then measured in vitro (Fig. 1A). As expected, IKK activity increased with 15 min and 30 min of TNF-α treatment in wild-type cells. In contrast to a recent report (35), no differences in either the kinetics or magnitude of IKK activity were detected in GSK-3β knockout MEFs. In addition, no alteration in TNF-induced IKK activity was detected in 293T cells in which GSK-3 activity was blocked by the pharmacological GSK-3 inhibitor SB216763 (K. A. Steinbrecher and A. S. Baldwin, unpublished data).

FIG. 1.

GSK-3β does not control IKK activity in mouse embryonic fibroblasts. (A) IKKβ was immunoprecipitated from TNF-α-treated (10 ng/ml) GSK-3β wild-type and null MEF extracts. Kinase activity on a GST-IκBα substrate was assayed by immunoblotting kinase reactions and incubating with a phospho-specific IκBα antibody. (B) MEF cell extracts were analyzed by Western blotting for phospho-IκBα, total IκBα, as well as β-tubulin as a loading control. (C) Immunoblotting of IKK target residues on p105 using p105 phospho-specific antibodies with TNF-α-stimulated GSK-3β wild-type and null whole-cell extracts. (D) Western analysis was used to determine cytokine-mediated p65 serine 536 phosphorylation. +/+, GSK-3β wild-type MEFs; −/−, GSK-3β null MEFs.

Based on the lack of altered IKK activity under conditions in which GSK-3β was absent or suppressed, we next investigated whether phosphorylation of IKK substrate proteins is affected by loss of GSK-3β activity. Phosphorylation of IκBα at serines 32 and 36, a critical event in cytokine-induced NF-κB activation, was evident in TNF-α-exposed GSK-3β null MEFs and was quantitatively and temporally very similar to that seen in wild-type cells (Fig. 1B). Furthermore, the expected decrease and resynthesis in IκBα protein following addition of TNF-α was nearly identical in MEFs lacking GSK-3β compared to wild-type MEFs.

Because IKK-mediated phosphorylation of p105 on serines 933 and 927 was suggested to be indirectly regulated by GSK-3β (9), we determined the phosphorylation status of p105 following TNF-α treatment. Cytokine stimulation of wild-type MEFs leads to phosphorylation of serines 933 and 927 of p105 and loss of GSK-3β did not suppress IKK-mediated p105 phosphorylation on either residue (Fig. 1C). This further suggests that IKK activity and its ability to interact with important downstream targets of the NF-κB pathway are not affected by decreased GSK-3 activity.

We included in our analysis the assessment of the phosphorylation status of the NF-κB subunit p65, an event associated with the transactivation potential of NF-κB. Although the kinase(s) that controls phosphorylation of serine 536 is not well established, defects in IKK function often manifest as decreased phosphorylation of this residue (26, 31). In wild-type fibroblasts, TNF-α stimulation results in enhanced phosphorylation of p65 at S536, as expected (Fig. 1D). Similarly, cells lacking GSK-3β did not show compromised induction of p65-S536 phosphorylation upon treatment with TNF-α. These data again strongly suggest that fibroblasts derived from mice lacking GSK-3β neither have significant defects in TNF-α-initiated signal transduction pathways that culminate in activation of the IKK complex nor have significant defects in the ability of IKK to associate with and phosphorylate its downstream target proteins. Since our results contrast with some studies regarding the function of GSK-3β in regulating NF-κB signal transduction, we next employed a different model system in which to assess the role of GSK-3 in IKK activation.

A nontransformed intestinal epithelial cell line (IEC-18) and a pharmacological GSK-3 inhibitor (SB216763, hereafter called SB21) were used to determine the effects of loss of total GSK-3 function on IKK activity. In order to accurately investigate the specific functional relationship between GSK-3β and NF-κB it was imperative to differentiate between the loss of GSK-3 activity and secondary effects on NF-κB via elevated β-catenin protein.

In many primary and nontransformed cells, blockade of GSK-3α/β activity with structurally unrelated inhibitors such as SB21 or lithium does not result in accumulation of β-catenin, suggesting one or more alternative means of regulating β-catenin in this context (24, 36). We observed no increases in β-catenin in IEC-18 cells following several hours of incubation with a proven effective dose of 30 μM SB21 (K. A. Steinbrecher and A. S. Baldwin, unpublished data). Loss of GSK-3-mediated phosphorylation of β-catenin demonstrates the efficacy of the SB21 treatment (Fig. 2A) and further supports use of these cells as a suitable system in which to investigate the β-catenin-independent effects of loss of GSK-3 activity on NF-κB.

FIG. 2.

Inhibition of GSK-3 activity does not effect IKK function. (A) IEC-18 cells were incubated with either DMSO or 30 μM SB21 for one hour and cellular proteins were extracted. Western analysis was used to determine phospho-β-catenin and total β-catenin as well as β-tubulin levels for confirmation of even loading. (B) Immunoblotting of total and phosphorylated IκBα proteins in whole cell extracts from TNF-α-treated (10 ng/ml) IEC-18 cells which had been pretreated with dimethyl sulfoxide (DMSO) or SB21 for one hour. (C) Western analysis showed phospho-p105 in dimethyl sulfoxide- or SB21-pretreated IEC-18 cells. (D) Phosphorylation of serine 536 on p65 was determined by immunoblotting IEC-18 extracts that had been treated as indicated.

We first determined the necessity of GSK-3 activity for proper IκBα phosphorylation and degradation. IEC-18 cells were treated with SB21 for 1 hour prior to addition of TNF-α. As in MEFs lacking GSK-3β, SB21-pretreated IEC-18 cells showed no differences from dimethyl sulfoxide-treated control cells in TNF-α-induced IκBα phosphorylation, degradation, and resynthesis (Fig. 2B). In addition, IEC-18 cells pretreated with SB21 and exposed to TNF-α demonstrated no defect in p105-S933 phosphorylation (Fig. 2C). Finally, cytokine-induced p65 phosphorylation at serine 536 was unaltered in GSK-3-blocked intestinal epithelial cells (Fig. 2D). Collectively, these data demonstrate that decreased GSK-3α and GSK-3β activity does not have major effects on the primary signaling pathways that mediate TNF-α-induced IKK activation and further suggest that the manner in which GSK-3β may regulate NF-κB activity occurs downstream of IκBα degradation.

GSK-3β is not required for NF-κB nuclear accumulation but is necessary for optimal DNA binding efficiency of NF-κB.

We next sought to determine the effects of loss of GSK-3β on signaling downstream of IKK activation and IκBα degradation. Coincident with the normal cytokine-elevated IKK activity and subsequent degradation of IκBα in cells is the release of NF-κB for accumulation in the nucleus. Therefore, we first addressed nuclear accumulation of p65 in GSK-3β wild-type and null MEFs. TNF-α stimulates the accumulation of p65 into the nucleus in similar amounts in wild-type and GSK-3β knockout MEFs (Fig. 3A). Additionally, levels of p65 decreased at a similar rate in the nuclei of both genotypes, suggesting no defects in either the rate of p65 nuclear entry or the rate of p65 export. Thus, similar levels of nuclear p65 are detected in both wild-type and GSK-3β null cells following cytokine stimulation.

FIG.3.

GSK-3β is required for efficient TNF-α-induced p65 DNA binding. (A) Nuclear extracts from TNF-α-treated (10 ng/ml) GSK-3β wild-type (+/+) and null (−/−) cells were analyzed by Western blotting using p65-specific antibodies. TFIIB immunoblotting was performed to confirm equivalent loading. (B) Electrophoretic gel shift assays were performed using nuclear extracts from TNF-α-stimulated cells. Antibodies that recognize either p65 or c-Rel were used in supershift analysis to identify shifted complexes in both GSK-3β wild-type and null cells. (C) DNA affinity pull-down assays were performed by incubating biotinylated DNA duplexes of NF-κB consensus site sequences with nuclear extracts from TNF-α-treated cells. Levels of bound p65 were determined by PAGE and immunoblotting. (D) DNA binding activity of p65-containing NF-κB dimers was determined using an ELISA-based approach. Nuclear extracts were incubated on ELISA plates harboring fixed NF-κB consensus site oligonucleotides and, following incubation with p65-specific and then horseradish peroxidase-conjugated antibodies, optical density in the presence of chromogenic substrate was measured. Gray bars indicate wild-type GSK-3β, and black bars indicate GSK-3β null cells (± standard error of the mean, n = 3). (E) NF-κB luciferase reporter assays were performed by TNF-α stimulation (6 h) of transfected GSK-3β wild-type and null MEFs (± standard error of the mean, n = 3).

The potential involvement of GSK-3β in mediating effective NF-κB DNA binding potential was explored with gel mobility shift assays. In wild-type cells the expected increase in shifted probe is seen at 15 and 30 min and the addition of supershifting antibodies reveals the major complex as a p65-containing NF-κB dimer (Fig. 3B). However, in GSK-3β knockout MEFs, there is a measurable decrease in shifted probe levels at all time points, although poor induction of DNA binding was most evident at 15 min.

DNA affinity pull-down assays were performed by incubating biotinylated NF-κB consensus site DNA duplexes with nuclear extracts from TNF-α-treated GSK-3β wild-type or null cells. Pull-downs with streptavidin-coated magnetic beads followed by Western blotting demonstrated that, in cells lacking GSK-3β, diminished amounts of p65-containing NF-κB dimers were able to bind to an NF-κB consensus site probe (Fig. 3C).

We further demonstrated the effects of loss of GSK-3β on NF-κB activity by using an ELISA-based assay to measure TNF-α-induced DNA binding. This approach detects a moderate yet still significant decrease in the amount of NF-κB bound to consensus site oligonucleotides fixed to the ELISA plate at both 15 and 30 min following TNF-α stimulation (Fig. 3D). We next investigated whether loss in DNA binding resulted in reduced induction of an NF-κB luciferase reporter by TNF-α. We found, as have others (19), that NF-κB luciferase reporter assays in GSK-3β wild-type and null cells demonstrated a significant decrease in reporter activity in response to TNF-α in cells lacking GSK-3β (Fig. 3E). In total, these data are consistent with the report indicating that GSK-3β is required for robust activity of NF-κB in mouse embryonic fibroblasts (19).

We next investigated TNF-α-induced NF-κB activity in IEC-18 cells, in which both GSK-3α and GSK-3β activity were inhibited by a pharmacological approach. Stimulation with TNF-α resulted in elevated levels of p65 in the nuclei of both dimethyl sulfoxide- and SB21-treated cells, a result that is consistent with normal IKK activity and IκBα degradation in cells lacking GSK-3 activity (Fig. 4A). As in GSK-3β null MEFs, the nuclear levels of p65 in SB21-pretreated cells was identical to controls. Gel shift mobility analysis was performed on nuclear extracts from untreated and SB21-treated IEC-18 cells. Dimethyl sulfoxide-treated cells showed the expected elevation of NF-κB DNA binding activity and the vast majority of this probe shift was due to p65-containing dimers (Fig. 4B, left). Similar to the results with MEFs lacking GSK-3β, we noted a substantial loss in DNA binding potential in cells pretreated with SB21 (Fig. 4B, left). Control gel shifts using an Oct-1 probe showed no deleterious effects on DNA binding by the SB21 compound (Fig. 4B, right).

FIG. 4.

Blockade of GSK-3 results in a loss of p65 DNA binding in TNF-α-treated cells. (A) IEC-18 cells were pretreated with either dimethyl sulfoxide (DMSO) or SB21 (30 μM) for 1 hour, stimulated with TNF-α (10 ng/ml), and nuclear proteins were isolated. Localization of p65 to the nucleus was determined by immunoblotting. (B) EMSA was performed as detailed in Fig. 3B and in Materials and Methods except that nuclear extracts from IEC-18 cells that had been pretreated with dimethyl sulfoxide or SB21 were used. DNA binding of Oct-1 to a radiolabeled probe in extracts from dimethyl sulfoxide- and SB21-treated cells (60 min) is shown at right. (C) DNA affinity pull-down assays were performed on control and GSK-3-blocked IEC-18 cells stimulated with TNF-α, and immunoblotting was used to determine levels of p65 which bound to a biotinylated duplex containing an NF-κB site. (D) ELISA-based measurement of p65 DNA binding was analyzed in IEC-18 cells as in Fig. 3D. Gray bars indicate dimethyl sulfoxide-treated IEC-18 extracts and black bars indicate SB21 pretreatment (± standard error of the mean, n = 3).

We next used the DNA affinity pull-down assay to further characterize the ability of TNF-α signaling to elevate NF-κB DNA binding in the presence of minimal GSK-3 activity. Biotinylated NF-κB consensus site duplexes were incubated with nuclear extracts from untreated or TNF-α-treated IEC-18 cells. Duplexes were then isolated, washed and the associated proteins were fractionated using PAGE. The expected increase in bound p65 was seen in dimethyl sulfoxide-treated control cells and yet immunoblotting of cells pretreated with SB21 demonstrated no increase in DNA binding of p65-containing NF-κB dimers (Fig. 4C). Measurement of NF-κB DNA binding activity was also determined using consensus site duplexes that were fixed to ELISA plates and demonstrated a similarly profound decrease in DNA-bound p65/NF-κB dimers (Fig. 4D). Taken together, data from MEFs and IEC-18 cells, which lack GSK-3β activity, support a role for this kinase in regulating the ability of NF-κB to efficiently bind DNA at certain target sequences. Furthermore, these data implicate GSK-3 proteins in controlling NF-κB-dependent gene transcription in a manner independent of β-catenin.

Expression of a subset of NF-κB-regulated genes is minimally effected by loss of GSK-3β.

Having established that GSK-3β does not regulate TNF-α signal transduction pathways that initiate IKK activation but that it does alter DNA binding activity of p65, we next sought to determine if loss of GSK-3β led to similarly broad, negative effects on NF-κB-mediated gene transcription. Our initial studies determined that, using relative quantitation by real-time reverse transcription-PCR, TNF-α-induced levels of IκBα mRNA are similar in both wild-type and GSK-3β null MEFs (Fig. 5A). This is consistent with the finding of no defects in IκBα protein resynthesis following TNF-α-induced proteosomal degradation (Fig. 1B) and (19).

FIG. 5.

Effects of GSK-3β on NF-κB DNA binding activity do not impact transcription of all NF-κB-dependent genes. (A) Real-time RT-PCR was performed on GSK-3β wild-type and null cells to assess changes in mRNA levels of IκBα relative to the wild-type, unstimulated time point. (B) Chromatin immunoprecipitation was used to assay localization of p65 to the promoter regions of the IκBα gene as described in Materials and Methods. PCRs with DNA from input as well as no-antibody (No Ab) control immunoprecipitations were also performed. (C) Relative real-time RT-PCR quantification of changes in MIP-2 mRNA levels in GSK-3β wild-type and null cells following TNF-α stimulation. (D) ChIP assays were used to investigate the presence of p65 at the MIP-2 promoter in wild-type and GSK-3β null MEFs. In this figure, gray bars indicate GSK-3β wild-type and GSK-3β null samples are represented by black bars (± standard error of the mean, n = 3; these data are representative of three to five separate experiments).

To further clarify the role of p65 in IκBα transcription within the context of GSK-3β ablation, we next determined whether the lack of p65 DNA binding as measured in broad terms by gel shift, DNA affinity pull-down assay and ELISA were reflected in association of p65 with the IκBα promoter following TNF-α treatment. Chromatin immunoprecipitation assays were performed in wild-type and GSK-3β null cells using antibodies specific for p65. Immunoprecipitation of p65-cross-linked DNA fragments followed by PCR with primers specific for the IκBα promoter region (spanning −265 to +10, where the transcriptional start site is +1) suggest that there is no defect in recruitment of p65 to the IκBα promoter following TNF-α stimulation in MEFs lacking GSK-3β (Fig. 5B). Similarly, real-time RT-PCR analysis of transcriptional induction of another NF-κB-regulated gene, MIP-2, showed only minimal effects of loss of GSK-3β (Fig. 5C). ChIP studies confirm that association of p65 with the promoter region of MIP-2 occurs with similar kinetics in GSK-3β null MEFs as in wild-type cells (Fig. 5D).

Additionally, acetylation of histone H4 at the IκBα and MIP-2 promoter regions, an indicator of open chromatin structure and a requirement for efficient gene activation, was also assessed by ChIP assay. We observed high levels of H4 acetylation at both the IκBα and Mip-2 promoters in wild-type and GSK-3β knockout MEFs and suggest that GSK-3β is not required for histone acetylation at these promoters (K. A. Steinbrecher and A. S. Baldwin, unpublished data). Our studies concerning the transcriptional regulation of the NF-κB target genes IκBα and MIP-2 consistently showed no defects in TNF-α-induced activation of these genes and, in fact, a small yet consistent elevation in mRNA induction was often seen (Fig. 5A and 5B).

We have identified and are investigating a number of genes that are significantly upregulated in response to TNF-α signaling in cells lacking GSK-3β compared to wild-type cells, implying another level of regulation involving repression of basal and cytokine-induced NF-κB-mediated gene transcription by GSK-3 proteins (K. A. Steinbrecher and A. S. Baldwin, unpublished data). Of note, this is supported by recent studies suggesting that GSK-3β can have a repressive effect on NF-κB activity (2, 6). Overall, these results indicate that reduction of NF-κB DNA binding as measured by gel shift analysis and other methods does not lead to loss of transcription of certain NF-κB-dependent genes.

Requirement of GSK-3β for cytokine-induced expression of certain NF-κB-regulated genes correlates with recruitment of p65 to chromatin.

Despite the aforementioned examples of NF-κB-target genes that are not greatly effected by genetic removal of GSK-3β, we speculated that the loss of NF-κB DNA binding activity that occurs in the absence of GSK-3 should negatively affect at least some genes that are dependent on NF-κB for transactivation. Our studies have identified several genes that are poorly induced in response to TNF-α in cells lacking GSK-3β, including baculoviral IAP repeat-containing 3 (commonly called cIAP-2), growth arrest, and DNA damage-inducible beta (GADD45β) as well as MCP-1 and IL-6. Here, we report mRNA expression levels and ChIP analysis of the latter two.

Expression of MCP-1 was strongly elevated by TNF-α treatment in wild-type MEFs but was minimally induced in cells lacking GSK-3β (Fig. 6A). ChIP studies demonstrate a substantial elevation in p65 at the MCP-1 promoter 30 min following TNF-α stimulation and moderate elevation at subsequent time points in wild-type cells (Fig. 6B). Consistent with poor MCP-1 mRNA expression found in GSK-3β knockout MEFs, p65 was recruited to the MCP-1 promoter at only very low levels following cytokine treatment (Fig. 6B). Quantitation of immunoprecipitated, p65-associated MCP-1 promoter DNA was performed using real-time PCR and confirmed the weak increase in promoter-bound p65 in cells lacking GSK-3β (Fig. 6C). Similar findings were observed regarding IL-6 gene activation. Elevation in IL-6 mRNA levels was completely absent in GSK-3β null MEFs upon TNF-α stimulation (Fig. 6D).

FIG. 6.

GSK-3β is required for TNF-α-induced transcription and promoter recruitment of p65 to a subset of NF-κB-regulated genes. (A) Real-time RT-PCR for MCP-1 expression was performed on wild-type and GSK-3β null MEFs. (B) ChIP assays to determine p65 localization to the MCP-1 promoter were performed on TNF-α-treated MEFs. The negative control and input are shown in the lower panels. (C) Real-time PCR was used to quantitate p65 ChIP DNA in GSK-3β wild-type (WT) and null (KO) MEFs. Data are normalized to levels of input DNA and presented as percent of input for each sample. (D) IL-6 mRNA levels following TNF-α stimulation were measured by real-time RT-PCR. (E) ChIP assays in GSK-3β wild-type and null MEFs were used to measure recruitment of p65 to the IL-6 promoter. In this figure, gray bars indicate GSK-3β wild-type and black bars represent GSK-3β null samples (± standard error of the mean, n = 3; these data are representative of 3-6 separate experiments).

We used ChIP analysis to determine if defects in association of p65 with the IL-6 promoter during cytokine signaling caused this. The nearly complete loss of IL-6 mRNA induction during TNF-α treatment was reflected in ChIP analysis of p65 at the IL-6 promoter, as p65 was almost totally absent from the IL-6 promoter in MEF cells that lack GSK-3β (Fig. 6E). We noted that acetylation of histone H4 at the MCP-1 and IL-6 promoters was not substantially affected by loss of GSK-3β (K. A. Steinbrecher and A. S. Baldwin, unpublished data). ChIP analysis of the cIAP-2 and GADD45β promoters confirms that GSK-3β is required for recruitment of p65 to these loci as well (K. A. Steinbrecher and A. S. Baldwin, unpublished data).

We next sought to confirm the necessity of GSK-3β for expression of IL-6 and MCP-1 in mouse embryonic fibroblasts. We transiently transfected wild-type (GSK-3β WT) or kinase-dead (GSK-3β KD) expression constructs into GSK-3β null cells (Fig. 7). In GSK-3β null cells, restoration of GSK-3β activity was able to impart responsiveness to TNF-α treatment and elevate levels of induced MCP-1 and IL-6. Expression of dominant-negative GSK-3β in MEFs lacking GSK-3β protein did not result in consistently elevated MCP-1 or IL-6 expression following exposure to TNF-α. Taken collectively, these data demonstrate the existence of a group of NF-κB-regulated genes that require functional GSK-3β for efficient expression.

FIG. 7.

Expression of GSK-3β restores TNF-α-induced elevation in MCP-1 and IL-6 mRNA. (A) Transient transfection of wild-type (WT) and kinase-dead (KD) GSK-3β into GSK-3β null MEFs was followed by stimulation with TNF-α and real-time RT-PCR to determine relative MCP-1 mRNA levels. (B) IL-6 mRNA expression was determined using the same approach. MCP-1 and IL-6 expression is represented by gray bars in wild-type cells and black bars in knockout cells (± standard error of the mean, n = 3; these data are representative of two or three separate experiments).

Furthermore, we show a decrease in DNA bound p65 at the promoters of these NF-κB target genes in the absence of GSK-3β. These studies suggest that GSK-3β influences the accessibility of certain NF-κB DNA binding sequences within the context of local chromatin structure and/or is critical for a specifically nuclear NF-κB function involving recruitment to the promoters regions of a subset of NF-κB-regulated genes.

DISCUSSION

The initial characterization of GSK-3β null mice and the associated defect in DNA binding of NF-κB dimers in fibroblasts from these mice has resulted in numerous attempts to determine the necessity for GSK-3β in proper NF-κB function. We have investigated the role of GSK-3β in TNF-α-induced signal transduction pathways that culminate in IKK activation. Based on our experiments, we show no defects in activation of IKK in cells lacking GSK-3β. Consistent with this finding, phosphorylation of several IKK substrates known to be required for efficient activation of NF-κB are not disrupted in cells lacking GSK-3β activity. We extend the initial report citing diminished DNA binding activity of NF-κB by using a variety of techniques to investigate the significance of loss of p65 DNA binding potential (19).

We measured the activation of a number of NF-κB target genes and have determined that GSK-3β is required for appropriate promoter localization of p65 during transcription of a subset of NF-κB-target genes. These data support a model in which, in fibroblasts and intestinal epithelial cells, GSK-3β affects the DNA binding activity of NF-κB through means independent of IKK activity and p65 nuclear localization. We speculate that GSK-3β may have direct effects on chromatin structure such that it facilitates accessibility of transcription factors such as NF-κB at the promoter regions of some genes. GSK-3β also likely has direct effects on p65 through phosphorylation and this may contribute to the low levels of this NF-κB subunit at the IL-6 and MCP-1 genes despite TNF-α stimulation (6, 33).

Our data are completely in agreement with the initial characterization of GSK-3β null MEFS from Hoeflich et al. (19). However, our data regarding the role of GSK-3β in TNF-α-induced cytoplasmic signal transduction pathways differ significantly from those of some recent reports and suggest cell type-specific functions for GSK-3β. Sanchez et al. have shown that GSK-3β blocks IKK activity via interaction with IKKγ and prevents accumulation of p65 in the nuclei of TNF-α-treated primary astrocytes (32). Using GSK-3β null MEF cells, Takada et al. report sweeping defects in activation of a number of cytoplasmic signaling intermediates and in retention of p65 in the cytoplasm upon TNF-α stimulation (35). We suggest that use of alternative means of blocking GSK-3 in a cell line which does not respond with accumulation of β-catenin is a critical additional model for use in determining the effects of GSK-3 on NF-κB function.

One possible mechanism by which GSK-3β controls NF-κB activity may be through direct phosphorylation of NF-κB (p65), with this modification effecting DNA binding activity or dimerization. Some studies provide support for the direct phosphorylation of p65 by GSK-3β (6, 33). A recent report implicated GSK-3β in phosphorylation of p65 at serine 468 and suggested that this was an inhibitory modification that, in coordination with protein phosphatases, regulated basal and cytokine-induced levels of p65 transactivation (6). Although not addressed specifically by Buss et al., it seems likely that this site is phosphorylated by GSK-3α also, as both GSK-3α and GSK-3β have the same substrate consensus sequence. Additional work is necessary to determine the functional relevance of p65-S468 phosphorylation during cytokine-stimulated gene transcription in vivo.

The initial report of p65 as a GSK-3β substrate did not directly identify serine 468 as a GSK-3β target residue (33). However, the study by Schwabe et al. established that, in vitro, GSK-3β is able to phosphorylate several sites in the transactivation domain of p65. The authors suggested the loss of GSK-3β-mediated phosphorylation of p65 results in TNF-α sensitivity in hepatocytes due to greatly diminished NF-κB transcriptional activity. This group also showed that lithium was able to stabilize β-catenin in cultured hepatocytes. When interpreted in light of more recent reports showing that increased levels of β-catenin can antagonize NF-κB activity (10, 11), it seems possible that the substantial elevation of β-catenin which occurred as a result of the use of lithium to block GSK-3 had inhibitory effects on NF-κB activity that were independent of GSK-3β.

The studies described above highlight an important point of consideration regarding the modulation of NF-κB activity by GSK-3 proteins. The role of GSK-3 in controlling NF-κB cannot be addressed in the context of elevated β-catenin due to the confounding effects that high levels of this protein have on NF-κB transactivation. Lithium and other GSK-3 inhibitors have been used widely in transformed cells to investigate the role of GSK-3 in NF-κB activity and yet, in many cell types, including MEFs, 293T, and many colon and breast cancer cell lines, these compounds cause a large elevation in β-catenin protein (10, 11, 33; K. A. Steinbrecher and A. S. Baldwin, unpublished data). Our studies employ model cell lines that do not respond to decreased GSK-3 activity with elevated β-catenin, allowing us to accurately determine the role of GSK-3 in NF-κB-mediated transcription. The lack of β-catenin stabilization may be due to compensation by GSK-3α or to the presence of GSK-3-independent mechanisms for initiating β-catenin degradation (21, 24). In addition, this is consistent with the lack of β-catenin accumulation in many primary and nontransformed cell types in vitro and in in vivo tissues when using pharmacological inhibitors of GSK-3 (7).

We speculate that NF-κB DNA binding is controlled by GSK-3 proteins in a dose responsive manner which is dependent on the amount of total GSK-3 activity present in the cell. This explains the significantly lower levels of NF-κB DNA binding seen in SB21-treated IEC-18 cells compared to GSK-3β null MEFs. We further suggest that, while the DNA binding potential of NF-κB is decreased as measured by gel shift and other assays, the in vivo effects of loss of GSK-3 activity on NF-κB-mediated gene transcription can be considered only on a gene-by-gene basis and may further differ according to cell type.

We have determined that while some NF-κB-target genes are only slightly affected (Fig. 5) and others seem to require GSK-3β for efficient transactivation (Fig. 6), there exist a number of genes that are greatly elevated by cytokine stimulation during low levels of GSK-3 activity (K. A. Steinbrecher and A. S. Baldwin, unpublished data). A decrease in GSK-3 activity may result in loss of p65-S468 phosphorylation and, therefore, potentiate NF-κB transactivation of some genes. This is consistent with a recent report that demonstrates elevated NF-κB reporter assay levels during decreased GSK-3β activity and further indicates that the predominant effect of loss of GSK-3 on NF-κB function is dependent on cell type and stimulus (2).

As suggested by others, it is possible that p65 residues other than S468 are also targeted by GSK-3 and phosphorylation at these sites is required to maintain p65 in a transcriptionally competent state that allows efficient transcription of some genes upon cytokine stimulation (6). Our initial efforts to determine if GSK-3β regulates interactions between p65 and coactivator proteins suggest that binding of CBP to p65 following TNF-α stimulation is unaffected by loss of GSK-3β (W. Wilson and A. S. Baldwin, unpublished data). Importantly, GSK-3 proteins have suppressive effects on numerous signaling intermediates and, therefore, complete loss of all GSK-3 activity most likely releases these signaling pathways and transcription factors for unrestrained activity and transactivation. Consequently, the effect of GSK-3 loss on NF-κB activity must be considered within this context.

It is clear from our studies that NF-κB-regulated gene transcription is controlled by GSK-3β in a highly promoter-specific manner. While a number of NF-κB target genes are not properly elevated in response to TNF-α, we found that IκBα and MIP-2 are largely unaffected by loss of GSK-3β. The maintenance of IκBα and MIP-2 expression despite loss of GSK-3β may be due to the presence of multiple NF-κB sites which act in a cooperative fashion at these promoters to ensure stable DNA binding (20). In addition, p65-containing NF-κB dimers that are present in GSK-3β null cells may retain affinity for the specific NF-κB binding site sequences that are present in the promoters of these genes. While our DNA binding assays were performed using a typical NF-κB consensus site sequence, we acknowledge that even minor base pair differences have been shown to affect gene activation (20). We speculate that cooperation with other transcription factors may also be critical for preserving IκBα and MIP-2 expression in the context of decreased GSK-3β activity.

A possibility that is not mutually exclusive to the regulation of NF-κB by direct GSK-3 phosphorylation is that GSK-3 activity regulates another signaling pathway or protein that is essential for efficient NF-κB DNA binding and transactivation. A candidate for this is AP-1, which is known to regulate both IL-6 and MCP-1 and to cooperatively mediate gene transcription with NF-κB (28, 29, 34, 37, 40). Studies report that interaction between NF-κB p65 and c-Fos, c-Jun, and JunD and blockade of AP-1 activity by expression of a dominant negative mutant results in suppressed NF-κB transactivation (29, 34). In hepatocytes, JunD was able to stimulate NF-κB activity and may interact directly with p65 (29). That some AP-1 subunits play a critical role in resistance to TNF-α-induced hepatocyte apoptosis (18, 39), as do GSK-3β and p65, is further support for the notion of GSK-3-mediated coordination between AP-1 and NF-κB in gene transcription. In addition, JNK activity is altered in cells which lack GSK-3β, further implicating AP-1 in downstream ramifications of decreased GSK-3 activity (25).

Sufficient GSK-3β activity is essential for NF-κB-mediated protection from TNF-α-induced apoptosis in some cell types (19, 33). We have found a subset of antiapoptosis genes (e.g., cIAP2) that are regulated by NF-κB and require functional GSK-3β for transcriptional induction. The loss of elevated cIAP-2 expression may explain, in part, the TNF-α sensitivity of hepatocytes lacking GSK-3β (19, 33, 38). Further work to determine whether loss of this and other GSK-3β/NF-κB-regulated antiapoptosis genes is the source of sensitivity to TNF-α-induced cell death is ongoing.