Phosphatidylinositol 3-Kinase as a Mediator of TNF-Induced NF-κB Activation¹

Tumor necrosis factor is a multifunctional cytokine that is produced primarily by activated macrophages (1, 2, 3). Originally characterized by its antitumor activity and by the cytotoxic response that it elicits from transformed cells, TNF is now known to conduct a wide range of biological effects (1, 2, 3). In addition to its effects on the growth, differentiation, and metabolism of normal cells, TNF also plays an important role in endotoxic shock, immunity, inflammation, antiviral responses, and diseases such as cachexia and non-insulin-dependent diabetes (1, 2, 3). The ability of TNF to induce this broad range of biological effects is attributable to its potent gene regulatory properties that depend, at least in part, on its ability to activate the transcription factor NF-κB. This has prompted much investigation into the nature of the signaling mechanisms that are triggered by TNF to activate NF-κB.

The activation of NF-κB is a tightly regulated process that involves its translocation from the cytoplasm to the nucleus where it binds to cognate DNA sequences (4). NF-κB is composed of homo- and heterodimers of members of the Rel family of transcription factors and is normally sequestered in the cytoplasm through its interaction with the I-κB (inhibitory of NF-κB)³ family of inhibitory proteins (4). In response to external stimuli, I-κB proteins undergo rapid phosphorylation on specific serine residues. Phosphorylation of I-κBα on serines 32 and 36 and of I-κBβ on serines 19 and 23 facilitates their ubiquitination on neighboring lysine residues, thereby targeting these proteins for rapid degradation by a proteosome (4). Following the degradation of I-κB, NF-κB is released and is free to translocate to the nucleus and to activate target genes. In addition to TNF, IL-1, phorbol esters, viruses, LPSs, and UV light are potent inducers of NF-κB (4).

TNF-induced signals for the activation of NF-κB are first transmitted by the TNF receptors across the plasma membrane and then relayed through specific cytoplasmic proteins. One of these cytoplasmic proteins, TNFR-associated death domain-containing protein (TRADD), is a protein adaptor that interacts specifically with TNFR type I and is required for TNF-mediated induction of NF-κB and apoptosis (5). TRADD interacts, in turn, with another adaptor protein, TRAF2 (TNFR-associated factor 2), which is dedicated to the activation of NF-κB and the c-Jun N-terminal kinase (JNK) pathway (6). Because IL-1 also employs a member of the TRAF family, TRAF6, to activate NF-κB (7), these TRAFs may be common mediators in cytokine-mediated NF-κB activation.

The recent identification of two protein kinases that induce NF-κB upon overexpression has furthered our understanding of how TNF and IL-1 activate NF-κB. One of these kinases, the NF-κB-inducing kinase (NIK), is closely related to the mitogen-activated protein kinase kinase kinases and physically interacts with TRAF2 (8). The second protein kinase, I-κB kinase α (IKKα), physically interacts with both NIK and I-κBα and phosphorylates I-κBα in vitro on serines 32 and 36 (9, 10). Phosphorylation of these residues is a prerequisite for the signal-induced ubiquitination and degradation of I-κB. IKKα may exist as part of a large 500- to 900-kDa complex, which is composed of several other polypeptides (11, 12). In addition to IKKα, inducible phosphorylation of I-κB may, in fact, be the cooperative effort of two other proteins, IKKβ (11, 12) and IKKγ (13), which are also part of the 500- to 900-kDa protein complex. Although they appear to have significant differences in their biochemical properties, all three proteins may play an important role in NF-κB and IKK activation and may even form part of a kinase cascade within the complex. The mechanisms by which IKKs and NIK are activated or recruited to the TRAFs are not clear and remain to be elucidated. Furthermore, the relationships of the TRAFs, NIK, and IKKs with other proteins such as TANK/ITRAF and IL-1R-associated kinase (IRAK) are important in the TNF (14) and IL-1 (15) signaling pathways, respectively, and are under investigation. There is also evidence that protein kinases other than the IKKs, such as mitogen-activated /extracellular signal-regulated protein kinase kinase 1 (MEKK1), a regulator of the JNK pathway, are also involved (16). This raises the possibility that the phosphorylation of I-κB in response to various inducers may be orchestrated by a network of kinases. In addition to I-κB, phosphorylation of NF-κB has also been reported (17). Because this phosphorylation is critical for NF-κB function, it would be of great interest to identify the protein kinases that phosphorylate the relevant sites.

Our recent studies on IL-1 signaling pathways revealed that phosphatidylinositol 3-kinase (PI 3-kinase) plays a major role in the ability of IL-1 to activate NF-κB and AP-1 (18). In addition, Sontag et al. (19) have shown that activation of NF-κB by the SV40 small t Ag requires PI 3-kinase and the atypical protein kinase C ζ (PKCζ) isoform. PI 3-kinase is a lipid kinase that is composed of two polypeptides, a p85 regulatory subunit, and a p110 catalytic subunit and is activated by a large spectrum of cytokines, hormones, and growth factors (20). It has been implicated as a key signaling molecule in processes as diverse as glucose transport (21), transcription factor activation (22), and cell survival (23, 24, 25, 26, 27). Our studies on IL-1 signal transduction implicating PI 3-kinase in the activation of NF-κB together with the report (28) on TNF-induced physical interaction between the insulin receptor substrate-1 (IRS-1) and PI 3-kinase suggest that this enzyme may also participate in TNF-induced activation of NF-κB. In this study, we present evidence that further supports a role for PI 3-kinase in TNF signaling.

HepG2 cells were obtained from the American Type Culture Collection (Manassas, VA) and maintained at 37°C in MEM containing 10% FBS and antibiotics. Wortmannin, insulin, and PI were from Sigma (St. Louis, MO). LY294002 was from Biomol (Plymouth Meeting, PA), IL-1 was obtained from the Biological Response Modifiers Program of the National Cancer Institute (Frederick, MD), and TNF was purchased from PeproTech (Rocky Hill, NJ). Anti-phosphotyrosine PY20 Abs were obtained from Transduction Labs (Lexington, KY). Anti-IRS-1 Abs were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Abs against the phosphorylated form of Akt kinase (serine 473) were from New England Biolabs (Beverly, MA).

HepG2 or U937 cells (5 × 10⁵) were seeded in 60-mm culture dishes and cultured overnight before they were serum starved for 4 h and treated with or without 1.4 nM TNF for the indicated times. Whole-cell lysates were prepared in ice-cold lysis buffer containing 50 mM HEPES (pH 7.5), 10% glycerol, 150 mM NaCl, 1% Nonidet P-40, 1.5 mM MgCl₂, 1 mM EGTA, 50 mM NaF, 1 mM sodium orthovanadate, 14 mM 2-ME, 0.2 mM PMSF, 10 μg/ml aprotinin, 10 μg/ml leupeptin, and 3 mM benzamidine. After 20 min on ice, the lysates were clarified by centrifugation at 14,000 × g. The supernatant was then incubated at 4°C with either PY20, anti-IRS-1, or anti-phosphoAkt Abs for 1 h. Immune complexes from extracts immunoprecipitated with PY20 or anti-IRS-1 Abs were then utilized for PI 3-kinase assays (29), and the labeled [³²P]PI 3-phosphate was extracted and resolved by thin-layer chromatography (30). Anti-phosphoAkt immune complexes were assayed for Akt kinase activities (25) using myelin basic protein (MBP) as substrate. The kinase incubation mixtures were then resolved on 10% SDS-PAGE gels and phosphorylated MBP was visualized by autoradiography.

Cells were treated with TNF and various concentrations of wortmannin. Nuclear extracts were prepared from treated cells and were incubated with a radiolabeled probe (2 × 10⁴ cpm) as described (18). Protein-DNA complexes were resolved on 5% polyacrylamide gels and visualized by autoradiography.

HepG2 (5 × 10⁵) cells were plated in 100-mm dishes and transfected the next day with the (NF-κB)₃/ chloramphenicol acetyltransferase (CAT) reporter gene and various expression vectors (18). Transfections were done using the Polybrene method (31). Sixteen hours after transfection, cells were treated with cytokines or left untreated for 20 h. All other procedures were performed as described (31).

U937 cells were treated for 90 min with 1.4 nM TNF or IL-1 in the presence of either 100 nM wortmannin or 10 μg/ml cycloheximide. Whereas cycloheximide was added at the same time as TNF or IL-1, wortmannin was added 45 min before the cytokines. In those experiments that required pretreatment, cells were incubated with 1.4 nM IL-1 or 1 μM insulin for 5 h with or without the inhibitors and then were treated with TNF plus either cycloheximide or wortmannin. After treatment for the indicated time periods, cells were immediately chilled on ice, harvested, and used for DNA (for DNA fragmentation analysis) or protein (for immunodetection of poly(ADP-ribose) polymerase (PARP)) extraction. For DNA extraction, cells were lysed in buffer containing 10 mM Tris-HCl (pH 8.0), 100 mM NaCl, 25 mM EDTA, and 0.5% SDS and incubated at 50°C for 10 min. Lysates were then incubated with ribonuclease at 37°C for 1 h, and then freshly prepared proteinase K (3 mg/ml) was added to digest protein at 37°C for 16 h. Fragmented DNA was resolved on 1.8% agarose gels at 35 V for 8 h and visualized by ethidium bromide staining. To detect PARP, nuclear extracts were prepared as described for EMSA, boiled in Laemmli buffer, and resolved on 7% SDS-polyacrylamide gels. Protein was transferred overnight to nitrocellulose membranes at 4°C. These membranes were then blocked with 5% nonfat dry milk before being incubated with anti-PARP mAb (PharMingen, San Diego, CA) for 60 min and then with enhanced chemiluminescence (Amersham, Arlington Heights, IL).

Green fluorescent protein (GFP) expression vector was transfected (Fugene 6, Boehringer Mannheim, Indianapolis, IN) into U937 cells with expression vector for p65 Rel A or with empty vector control at a ratio of 1:3. After 20 h, cells were treated with various combinations of TNF, cycloheximide, or wortmannin. Samples were then coded and processed in a double-blinded fashion. After treatment, cells were harvested, stained with Hoechst 33342, and observed under a fluorescent microscope. Photographs of GFP and Hoechst-stained cells were taken of randomly selected fields in each treatment. For statistical analysis, the total number of GFP-expressing cells were counted along with the number of GFP-positive cells that were nonapoptotic as determined by Hoechst staining. Results from three independent experiments were used to estimate the cell viability and to calculate the SEs.

In our previous studies on the signaling pathways triggered by IL-1, we found that PI 3-kinase was a necessary component of IL-1-induced NF-κB activation (18). Because TNF is the other major cytokine that can potently activate NF-κB, we sought to evaluate the possibility that PI 3-kinase might also be an integral part of TNF signaling. We first examined whether PI 3-kinase activity and its downstream target, Akt kinase, could be induced in response to TNF stimulation. HepG2 cells were treated with TNF for various periods of time and used to prepare whole cell extracts for immunoprecipitation of PI 3-kinase and phosphorylated Akt kinase activities. Two different Abs were used to immunoprecipitate PI 3-kinase activity from these extracts: 1) anti-phosphotyrosine Abs (PY20), which are routinely used to detect changes in PI 3-kinase activity after treatment with various agents (see references in Ref. 20), and 2) anti-IRS-1 Abs, because IRS-1 has been reported to interact with PI 3-kinase upon TNF stimulation (28). Abs against phosphorylated Akt (serine 473) were used to immunoprecipitate Akt kinase activity. As shown in Fig. 1,A, there was a rapid but transient increase in PI 3-kinase activity after treatment with TNF. Within 0.5 min there was an ∼3-fold activation of the enzyme, and maximum activation (4- to 7-fold) was observed within 1 min of stimulation. Although the extent of PI 3-kinase activation detected with anti-IRS-1 Abs was lower than that observed with PY20, the kinetics of activation observed with either Ab were nearly identical. To ensure that the induction of PI 3-kinase activity by TNF is not unique to HepG2 cells, U937 cells were also stimulated with TNF and the effect on PI 3-kinase activity was determined. Indeed, addition of TNF also stimulated PI 3-kinase activity in U937 cells with induction kinetics similar to those in HepG2 cells (Fig. 1,A). Consistent with its function in the PI 3-kinase signaling cascade, Akt kinase activity was also increased (∼3-fold in 20 min; Fig. 1 B) in response to TNF. The magnitude and kinetics of induction of Akt kinase by TNF are comparable to those observed for insulin (30). This further supports a functional role of the PI 3-kinase pathway in TNF signal transduction.

Our results are consistent with the observations of Guo and Donner (28) that TNF induces a physical association between IRS-1 and PI 3-kinase and further suggest that this interaction may serve as a mechanism for the activation of PI 3-kinase. Physical interaction between PI 3-kinase and IRS-1 may involve interactions between the SH2 domains of the p85 regulatory subunit of PI 3-kinase and the phosphotyrosine-containing YXXM motifs present on IRS-1 (20). In this regard, it is noteworthy that no potential PI 3-kinase binding motifs were found in the protein sequences of any of the known TNF signaling molecules recruited for the activation of NF-κB. Therefore, it is likely that at least one of these signaling proteins, such as the TNF receptors, interacts indirectly with PI 3-kinase through IRS-1.

To identify the role of PI 3-kinase in TNF signaling, we used various inhibitors to see whether they would block the activation of known transcription factors by TNF. First, we tested the effect of wortmannin, a selective PI 3-kinase inhibitor, on the ability of TNF to increase the DNA-binding activity of NF-κB. This activity of TNF is based on its ability to induce the degradation of the NF-κB inhibitor, I-κB, and is recognized as a critical step before the transcription factor can be translocated to the nucleus. Wortmannin had little or no effect on TNF-induced NF-κB-binding activity (Fig. 1 C) even at concentrations of 400 nM, suggesting that PI 3-kinase is not involved in the signaling pathway leading to the degradation of I-κB.

Because I-κB degradation alone does not fully account for the functional activation of NF-κB-dependent genes (17), we used transient cotransfection assays to examine whether PI 3-kinase was necessary to transactivate NF-κB-dependent reporter gene expression. An (NF-κB)₃/CAT reporter gene that contained three copies of the NF-κB-binding site was cotransfected into HepG2 cells with an expression vector encoding either the dominant-negative mutant of the p85 subunit of PI 3-kinase (p85DN) (21) or the catalytic p110 subunit (20). The dominant-negative mutant p85DN is identical with the wild-type p85 except that it lacks the sequence essential for its association with the p110 subunit (21). Consequently, p85DN is incapable of binding to p110, though it still can compete with wild-type p85 for specific PI 3-kinase-binding sites. This means that, relative to wild-type p85, p85DN would be competitively recruited by various PI 3-kinase-binding proteins such as IRS-1 but would be defective in transmitting signals downstream. As such, p85DN has been used frequently to demonstrate a role for PI 3-kinase in various cellular processes (21, 32) including SV40 small t Ag-induced NF-κB activation (19).

Because wortmannin had not blocked the activation of NF-κB-binding activity, it was surprising when overexpression of p85DN in HepG2 cells inhibited the TNF-mediated induction of the (NF-κB)₃/CAT reporter gene by about 80% (Fig. 2,A). These differential effects of PI 3-kinase inhibitors suggest that NF-κB is regulated at different steps through parallel signaling pathways. One pathway leads to the degradation of I-κB and the other may lead to an increase in the transactivation potential of NF-κB. Therefore, the effect of p85DN implicates PI 3-kinase in TNF-induced transactivation of NF-κB-dependent genes. The inhibition was dose-dependent and at a level comparable to that observed with IL-1. These inhibitory effects were deemed to be specific because in control experiments p85DN had no effect on the induction of an estrogen-responsive reporter gene by estradiol (Fig. 2 B) or on the induction of a T1 kininogen gene by the combination of IL-6 and dexamethasone (18).

Because PI 3-kinase appears to be necessary for TNF to induce NF-κB, we examined the possibility that overexpression of the catalytic subunit of this enzyme would activate NF-κB-dependent gene expression. However, cotransfection of the (NF-κB)₃/CAT reporter gene with p110 did not result in its induction (Fig. 2 C). This is consistent with the fact that several inducers of PI 3-kinase, such as hepatocyte growth factor and erythropoietin, are unable to activate NF-κB.

We next investigated the effect of TNF on p110-overexpressing cells because we had previously observed synergism between IL-1 and PI 3-kinase (18). When p110-overexpressing cells were stimulated with TNF, there was a dramatic induction of the (NF-κB)₃/CAT reporter gene (Fig. 2,C), signifying synergism between PI 3-kinase and TNF. To verify that this effect was mediated specifically through NF-κB, we cotransfected I-κBα-expression plasmids with the CAT reporter gene. As shown in Fig. 2 C, the induction of the CAT reporter by TNF and by the combination of TNF and p110 were both inhibited by the coexpression of I-κBα.

That PI 3-kinase was apparently required but insufficient to activate NF-κB suggests that PI 3-kinase may need to cooperate with another TNF-inducible signal(s) to activate NF-κB. This notion is consistent with our observation that p110 synergized with TNF to induce NF-κB-dependent gene expression. This mechanism of NF-κB activation may be similar for both TNF and IL-1 because p110 synergizes with IL-1 and may involve the cooperation ofp110 with IRAK (18). It is presently unclear whether PI 3-kinase would cooperate with any of the known transducers in the TNF signaling pathway.

Upon binding to its receptor, TNF initiates two very different signals that diverge downstream of the TNF receptor-interacting protein TRADD (5). One signal leads to the activation of NF-κB, whereas the other initiates apoptosis or programmed cell death. TNF-induced apoptosis is potentiated in the presence of cycloheximide, presumably due to inhibition of de novo synthesis of TNF-induced gene products that function to protect cells from programmed cell death (5). It has been demonstrated recently that these antiapoptotic genes are downstream targets of NF-κB; that is, constitutive activation of NF-κB can protect cells from apoptosis while inhibition of its activation potentiates TNF-induced apoptosis (5, 33, 34, 35).

Because we showed that TNF-mediated activation of NF-κB-dependent reporter gene requires PI 3-kinase, we sought to examine the effects of wortmannin treatment on TNF-induced apoptosis through its inhibition of NF-κB transactivation activity. U937 cells were treated with TNF alone or in combination with wortmannin or LY294002. To monitor the extent of apoptosis, DNA fragmentation assays and immunodetection of the apoptosis-induced cleavage of PARP were performed on treated cells. As controls, U937 cells were treated with the combination of TNF and cycloheximide. Consistent with earlier reports, neither TNF nor cycloheximide alone could induce apoptosis (Fig. 3, A and B, lanes 2 and 3), whereas treatment with TNF plus cycloheximide resulted in extensive DNA fragmentation and PARP cleavage (Fig. 3, A and B, lane 4). When wortmannin was added together with TNF, it also strongly potentiated TNF-induced apoptosis to a level comparable to that observed with cycloheximide (Fig. 3, A and B, lanes 4 and 5). This is consistent with the ability of TNF to activate PI 3-kinase in these cells (Fig. 1,A). No apoptosis could be detected in control cells that were treated with wortmannin alone. LY294002, a more specific PI 3-kinase inhibitor than wortmannin, also potentiated TNF-induced PARP degradation (Fig. 3 B). This data indicates that both inhibitors can potentiate TNF-mediated apoptosis by specific inhibition of PI 3-kinase activities. Furthermore, IL-1, which is not known to induce apoptosis, had no effect either alone or in combination with cycloheximide or wortmannin in these experiments. Therefore, these data are consistent with the involvement of PI 3-kinase in TNF-induced activation of NF-κB function. It is noteworthy that although cycloheximide, wortmannin, and LY294002 can potentiate the ability of TNF to induce apoptosis, their sites of action differ: whereas cycloheximide blocks events downstream of the activation of NF-κB-dependent genes, wortmannin and LY294002 inhibit signaling events that are upstream to the activation of those genes.

Cells that are normally susceptible to TNF/cycloheximide-induced apoptosis can be protected by pretreatment with IL-1 due to its ability to activate NF-κB (35). Therefore, we examined whether pretreatment with IL-1 could also protect cells from TNF/wortmannin-induced apoptosis. Because both IL-1 and TNF can induce PI 3-kinase activity (Ref. 18 and Fig. 1) and PI 3-kinase activity has been shown to protect cells from serum withdrawal-, UV-, and myc-induced apoptosis (23, 24, 25, 26, 27), we included insulin pretreatment as a control to differentiate the NF-κB effects from those of PI 3-kinase. Cells pretreated for 5 h with IL-1 were effectively protected from subsequent TNF/cycloheximide- or TNF/wortmannin-induced apoptosis (Fig. 3,C, lanes 2 and 5). The protection conferred by IL-1 could be blocked by the simultaneous treatment with wortmannin to inhibit the IL-1 signaling pathway (Fig. 3,C, lanes 3 and 4). At the concentrations employed, wortmannin alone did not induce apoptosis even after 7 h (see below). However, pretreatment with insulin did not confer any protection (Fig. 3,C, lanes 6 and 7) despite its ability to activate PI 3-kinase in these cells (36) and to protect PC-12 cells from serum withdrawal-induced apoptosis (23). This suggests that PI 3-kinase-dependent survival signals, although sufficient to protect cells from serum withdrawal- and UV-induced apoptosis, may not be sufficient to protect cells from TNF-induced apoptosis. Because both IL-1 and insulin activate PI 3-kinase but only IL-1 can also activate NF-κB, these data also suggest that apoptosis induced in cells by treatment with TNF and wortmannin is due to inhibition of NF-κB activation. Furthermore, the involvement of PI 3-kinase in NF-κB activation is also supported by the observation that wortmannin can inhibit the protection conferred by IL-1 (Fig. 3,C). To provide further evidence that potentiation of TNF-induced apoptosis by wortmannin is due mainly to inhibition of NF-κB rather than to the inhibition of other PI 3-kinase-associated survival pathways, we examined whether overexpression of NF-κB could protect cells from apoptosis induced by the combination of TNF and wortmannin. U937 cells were transfected with an expression vector for p65 Rel A or with an empty vector. To mark the transfected cells, GFP-expression vector was cotransfected with Rel A vector or the empty vector control at a ratio of 1:3. Approximately 20 h after transfection, cells were treated with various combinations of TNF, cycloheximide, and wortmannin, and the extent of apoptosis was analyzed after Hoechst staining. Treatment of U937 cells with TNF, cycloheximide, or wortmannin individually had minimal effects on cell survival even after 5–7 h of treatment (Fig. 4,A, panels 5–12 and panels 17–20, and B). In contrast, the combination of TNF and cycloheximide reduced cell viability to almost 12% (Fig. 4 B). However, overexpression of p65/Rel A in these cells significantly increased the viability to 26%. These results are consistent with earlier reports that cycloheximide potentiates TNF-induced apoptosis by blocking NF-κB-inducible gene expression. In our case, we also observed protection from the combined effects of TNF and cycloheximide in p65/Rel A-transfected cells, but not in control cells.

Whereas treatment with wortmannin alone did not induce any significant apoptosis (Fig. 4), it potentiated TNF-induced apoptosis resulting in reduced cell viability to about 35% (Fig. 4,A, panels 21 and 22, and B). However, when these cells were transfected with p65/Rel A, they showed considerable resistance to apoptosis induced by this combination of TNF and wortmannin (Fig. 4 A, panels 23 and 24, and B). The protection from TNF/cycloheximide- and TNF/wortmannin-mediated apoptosis by p65/Rel A are highly significant, as indicated by their p values. If NF-κB activation in the TNF signaling pathway had not been interfered with, overexpression of p65/Rel A would not have been sufficient to protect these cells from apoptosis (33, 34, 35). Therefore, the fact that p65/Rel A overexpression can protect cells from TNF/wortmannin-induced apoptosis suggested that wortmannin potentiates TNF-induced apoptosis largely by inhibiting NF-κB activity. Therefore, taken together, our data are consistent with the involvement of PI 3-kinase in TNF-induced NF-κB activation. However, we cannot completely rule out a significant role for PI 3-kinase-mediated, NF-κB-independent survival pathways under our experimental conditions.

We have used a dominant-negative mutant of p85 PI 3-kinase to assess the involvement of PI 3-kinase in TNF-mediated activation of NF-κB. In this paper and in a previous report (18), we have provided evidence that this enzyme may be common to the IL-1 and TNF pathways leading to the activation of NF-κB, albeit may be through different mechanisms. We have also shown here that PI 3-kinase alone is insufficient to activate NF-κB but that it may interact with other mediators of TNF signal transduction because it greatly synergized with TNF. The other mediators of TNF signal transduction that PI 3-kinase interacts with to activate NF-κB are uncertain, but such interactions appear to be essential. Therefore, this is reminiscent of the IL-1 signal-transduction pathway in which NF-κB activation may be dependent, at least in part, on the interaction between PI 3-kinase and the serine/threonine protein kinase IRAK (18). Whether the synergism between PI 3-kinase and IRAK is mediated by common mediators of the IL-1 and TNF signaling pathways, such as NIK, or by the IKK kinase complex remains to be determined. Functional interaction of PI 3-kinase with specific TNF-inducible signals may contribute to superactivate downstream signaling proteins, to target them appropriately to subcellular compartments, or to relieve tight negative regulation.

PI 3-kinase may not, in fact, participate in any linear pathway that leads to NF-κB activation in as much as several known PI 3-kinase activators such as HGF cannot induce NF-κB. Given that overexpression of proteins such as TRAF2 and TRAF6 can activate NF-κB, it would appear possible that a linear pathway for NF-κB activation may be functional in the absence of PI 3-kinase. However, without PI 3-kinase, the kinetics and/or the extent of NF-κB activation or transactivation potential may be severely compromised. We favor the last possibility, because wortmannin only minimally inhibited (if at all) the DNA-binding activity of TNF-induced NFκB in gel-mobility shift assays, whereas expression of the dominant-negative inhibitor p85DN blocked TNF-induced NF-κB-dependent gene expression by about 80%. The inhibitory effects of p85DN on the induction of NF-κB-dependent genes by TNF suggest that PI 3-kinase may be involved in a step or a series of steps that regulate the transactivation potential of NF-κB but not its DNA-binding. Our data suggest that while PI 3-kinase may participate in modulating increases in the transactivation potential of NF-κB, it appears not to be involved in TNF-induced I-κB degradation. A similar role for PI 3-kinase has been postulated in the activation of STAT3 (22). Tyrosine phosphorylation of STAT3 is sufficient for a profound activation of this transcription factor; however, phosphorylation of STAT3 on a specific serine residue mediated by PI 3-kinase is necessary for it to achieve full activity.

Two recent reports by Ozes et al. (37) and Romashkova and Makarov (38) have implicated PI 3-kinase/Akt kinase signaling pathway in NF-κB activation. These studies showed that both platelet-derived growth factor and TNF-mediated NF-κB activation involve the activation of Akt kinase and its interaction with the IKKs. However, they disagree on whether TNF-induced NF-κB activation involves the activation of PI 3-kinase. Although Romashkova and Makarov (38) failed to observe any inhibitory effects of wortmannin on the induction of a NF-κB reporter gene by TNF, Ozes et al. (37) reported that wortmannin, a dominant-negative p85 PI 3-kinase mutant, and a kinase deficient Akt kinase are all able to abrogate TNF-induced activation of NF-κB. Although our data indicate a role for the PI 3-kinase pathway in TNF-mediated NF-κB activation, there are mechanistic differences between the model we propose and that suggested by the observation of Ozes et al. (37). Indeed, because we did not observe any inhibitory effects of wortmannin on the ability of TNF to increase the DNA-binding activities of NF-κB, our data would suggest that PI 3-kinase may be dispensable for I-κB degradation. These differences, while they cannot be fully explained, may suggest that PI 3-kinase affects NF-κB activation in more than one way and that the nature of its effects may vary in a cell type-specific manner.

The mechanism of action of PI 3-kinase in NF-κB activation may also involve its phosphorylated lipid products. These lipid mediators may regulate the activity of protein kinases such as PKCζ (39), an atypical PKC isoform that has been implicated in NF-κB activation (40). These protein kinases may then regulate signaling proteins involved in TNF signal transduction. This possibility is also supported by recent observations that SV40 small t Ag-mediated NF-κB activation is dependent on both PKCζ and PI 3-kinase (19). These studies also indicated that PKCζ is a downstream target of PI 3-kinase in vivo. The small G protein Cdc42 (41) and the protein kinase MEKK1 (16) are some of the other proteins implicated in TNF-induced activation of NF-κB. Because physical interaction between Cdc42 and PI 3-kinase (42) may result in the stimulation of PI 3-kinase activity (43), it remains to be seen whether these interactions are important for the activation of NF-κB. MEKK1, a key participant in the activation of JNK, has been implicated in the activation of the 700-kDa I-κB kinase complex (16). A signaling link between MEKK1 and PI 3-kinase seems possible because overexpression of PI 3-kinase results in JNK activation (44). To understand the nature of the involvement of PI 3-kinase in NF-κB activation, it is important to identify its relationship with the other signaling molecules recruited by TNF.

We thank Drs. M. Kasuga and W. Ogawa for the p85DN construct; A. Klippel and L. T. Williams for the p110α expression vector; C. Reynolds and the Biological Response Modifiers Program, National Cancer Institute, for IL-1β; and P. J. Chiao for the I-κBα expression vector. We are grateful to Karen Hensley for help with the figures and to Zhanyong Bing for help with statistical analysis. We are also very thankful to Drs. Athula Wikramanayake and Masahiro Higuchi for help with the immunofluorescence experiments.