Nuclear Factor kB-independent Cytoprotective Pathways Originating at Tumor Necrosis Factor Receptor-associated Factor 2*

Most normal and neoplastic cell types are resistant to tumor necrosis factor (TNF) cytotoxicity unless cotreated with protein or RNA synthesis inhibitors, such as cycloheximide and actinomycin D. Cellular resistance to TNF requires TNF receptor-associated factor 2 (TRAF2), which has been hypothesized to act mainly by mediating activation of the transcription factors nuclear factor kB (NFkB) and activator protein 1 (AP1). NFkB was proposed to switch on transcription of yet unidentified anti-apoptotic genes. To test the possible existence of NFkB-independent cytoprotective pathways, we systematically compared selective trans-dominant inhibitors of the NFkB pathway with inhibitors of TRAF2 signaling for their effect on TNF cytotoxicity. Blockade of TRAF2 function(s) by signaling-deficient oligomerization partners or by molecules affecting TRAF2 recruitment to the TNF receptor 1 complex completely abrogated the cytoprotective response. Conversely, sensitization to TNF cytotoxicity induced by a selective NFkB blockade affected only a fraction of TNF-treated cells in an apparently stochastic manner. No cytoprotective role for c-Jun amino-terminal kinases/stress-activated protein kinases (JNKs/SAPKs), which are activated by TRAF2 and contribute to stimulation of activator protein 1 activity, could be demonstrated in the cellular systems tested. Although required for cytoprotection, TRAF2 is not sufficient to protect cells from TNF + cycloheximide cytotoxicity when overexpressed in transfected cells, thus indicating an essential role of additional TNF receptor 1 complex components in the cytoprotective response. Our results indicate that TNF-induced cytoprotection is a complex function requiring the integration of multiple signal transduction pathways.

Intracellular signal transduction from p55 tumor necrosis factor (TNF) 1 receptor 1, which is the main TNF receptor on most cell types, occurs through a controlled series of protein-protein interactions (1)(2)(3). Following TNF-induced trimerization of the receptor (4), TNF receptor-associated death domaincontaining protein (TRADD) (5) is recruited to a region of TNF-R1 to which the cytotoxic function has previously been mapped, namely the death domain (6 -8). TRADD acts as an adapter required to recruit to the receptor the downstream transducers Fas-associated death domain-containing protein (FADD) (9 -11), TNF receptor-associated factor 2 (TRAF2) (12) and receptor-interacting protein (RIP) (13,14). Whereas FADD interacts with and activates the apoptotic proteases (15,16), thus triggering cell death, TRAF2 has been implicated in the activation of two distinct pathways, one of which (17), which requires the recently identified protein kinase NFkB-inducing kinase (NIK) (18), activates a multiprotein catalytic complex (IkB kinase complex) (19 -24) that phosphorylates the NFkB inhibitory subunit IkB␣ at serines 32 and 36. Phospho-IkB␣ is then degraded, thus liberating NFkB, which enters the nucleus and activates transcription of target genes. The other pathway, which is independent of NIK (25,26), activates the c-Jun amino-terminal kinases/stress-activated protein kinases (JNKs/ SAPKs) (27)(28)(29), a family of extracellular signals-regulated Ser/Thr protein kinases that stimulate transcription by phosphorylation and activation of a number of transcription factors, including c-Jun, activating transcription factor 2 (ATF-2), and ternary complex factor (TCF)/Elk2 (30). Although its specific mode of action and function are not completely clear, the protein kinase receptor-interacting protein (14) may represent a critical component of the NFkB signaling apparatus, as suggested by the defective NFkB activation of a receptor-interacting protein-deficient mutant Jurkat cell line (31).
Therefore, although TNF-R1-induced apoptosis requires FADD, stimulation of gene expression mainly occurs through TRAF2-dependent pathways. The first biological role of TNFstimulated gene expression is to protect cells from TNF cytotoxicity: indeed, most TNF-treated cells are resistant to TNF cytotoxicity unless treated with protein or RNA synthesis inhibitors, such as cycloheximide or actinomycin D, respectively. Moreover, TNF pretreatment usually protects cells from a subsequent challenge with TNF ϩ cycloheximide (32). Evidence for a TRAF2 role in turning on protective genes originally came from experiments with a signaling-deficient TRAF2 mutant, TRAF2(87-501), that acts in a dominant negative fashion (17): unlike wild type cells, those expressing TRAF2(87-501) are almost invariably doomed to death upon TNF-R1 cross-linking (27,28). Moreover, degradation of endogenous TRAF2 by constitutive signaling through CD30, a TNF receptor superfamily member that also signals through TRAF2, is associated with increased cellular sensitivity to TNF (33). One critical downstream effector of TRAF2-dependent cytoprotection is transcription factor NFkB, as suggested by TNF␣ sensitivity of both cells lacking the p65/RelA NFkB subunit (34) and cells expressing a phosphorylation-defective IkB␣ mutant (IkB␣S32A/ S36A) (27,35,36), that, being unresponsive to extracellular signals, irreversibly sequesters NFkB in the cytoplasm, thereby acting as a "super-repressor" of NFkB function (37)(38)(39). Although the experimental evidence for a NFkB role in transcriptional induction of protective genes is strong, it is not yet clear whether additional TNF-R1-originating pathways do exist that contribute to cytoprotection against TNF-induced apoptosis.
In this study, we comparatively analyzed the effect of a selective NFkB inhibition with that of a complete TRAF2 blockade; the experiments were carried out in cellular systems in which TNF-induced new gene expression is both sufficient and required for protection against TNF-induced apoptosis, i.e. cells that are killed by TNF only in the presence of translational inhibitors and that when pretreated with TNF are protected against a subsequent challenge with TNF ϩ cycloheximide (CHX). Our results indicate that NFkB-independent TRAF2dependent pathways make a significant contribution to the TNF-induced cytoprotective response.
Cell Culture, DNA Transfections, NFkB Assays, and ␤-Galactosidase Assays-Hela and Chang cells were kept in standard culture conditions. 1.8 ϫ 10 5 cells were plated in 35-mm plates and transfected the next day with the indicated amount of DNA using a cationic lipid-based reagent (Superfect, Qiagen). NFkB assays and ␤gal assays were performed as described (25,28).
Infection of Primary Rat Oligodendrocytes with Recombinant Retroviruses-pBabe vectors were transfected in packaging cells, and secreted retrovirus was filtered and stocked (44). Primary rat oligodendrocytes cultures were obtained as described (45,46). Purified oligodendrocytes precursors were obtained by mechanical dissociation from rat mixed glial primary cultures (45,46). To prevent irreversible exit from cell cycle, cells were grown in chemically defined serum-free medium lacking triiodothyronine and thyroxine (which stimulate differentiation of precursors into oligodendrocytes) and containing platelet-derived growth factor and basic fibroblast growth factor (10 ng/ml). 24 h after plating, cells were restimulated with platelet-derived growth factor and basic fibroblast growth factor and simultaneously incubated with the indicated retrovirus for 36 h. Cells were then selected in puromycin (1 g/ml) for 4 days in serum-free differentiation medium. Before TNF treatment, cells were incubated with interferon ␥ (100 g/ml), which increases TNF-R1 expression and renders cells responsive to TNF stimulation. 2 After a 6-h TNF treatment (50 ng/ml), cells were fixed, permeabilized, and sequentially stained with anti-epitope antibodies and with terminal deoxynucleotidyltransferase-mediated dUTP nick-end labeling using a commercially available kit (Boehringer Mannheim).

Selective Inhibition of TNF-induced NFkB Activation by A20
Zinc Finger Protein-A20, a zinc finger protein that is expressed at very low levels in most cell types, is rapidly induced upon NFkB activation by TNF-␣, interleukin 1, CD40, phorbol myristate acetate, Epstein-Barr virus latent infection membrane protein 1, and human T-cell lymphotrophic virus 1 tax (48 -52). Although originally characterized as an inhibitor of TNF-induced cell death in stably expressing clones (53), A20 has later been shown to act as a potent inhibitor of NFkB activation by TNF, as well as by phorbol myristate acetate, oxygen radicals, and lipopolysaccharide (54 -57). We first characterized A20 effects on TNF-R1 signal transduction pathways in HeLa cells, a human epithelial cell line that undergoes TNF-dependent apoptosis only in the presence of translational inhibitors. Treatment of HeLa cells with TNF ϩ CHX caused massive cell death that was entirely dependent on FADD and on the subsequent activation of cystein-aspartate proteases (caspases), as shown by the complete inhibitory effect of both a dominant negative FADD mutant (11) and of the cowpox virusencoded caspase inhibitor CrmA (40) (Fig. 1a). High levels of exogenously overexpressed A20 did not affect apoptosis by either TNF ϩ CHX treatment or by TRADD overexpression (which induces apoptosis in a CHX-independent manner) ( Fig.  1a), thus indicating that the pro-apoptotic caspase pathway is fully active in A20-expressing HeLa cells. Conversely, A20 expression strongly inhibited both basal and TNF-stimulated NFkB activity (17) (Fig. 1b). In a similar manner, A20 strongly interfered with NFkB activation by both TRAF6 and phorbol myristate acetate but not p65/RelA (Fig. 1b). Although the molecular bases for NFkB inhibition by A20 are not clear, the fact that A20 is a broad spectrum (but not general) NFkB inhibitor preventing IkB␣ degradation 3 suggests that it probably acts downstream of TRAF2 and before IkB kinase along the NFkB pathway.
Inhibition of NFkB Activation by A20 Correlates with Sensitization to TNF-induced Apoptosis-Because A20 selectively inhibits NFkB activation by TNF-R1/TRAF2 without affecting the apoptotic pathway, and because NFkB has been shown to be required for cytoprotection from TNF-induced cytotoxicity, we would expect an increased rate of TNF-induced apoptosis in A20-expressing cells with respect to vector-transfected cells.
Indeed, transient A20 expression increased the percentage of TNF-induced death about 3-fold over the basal level (from about 10% to 25-30%). However, most transfected cells remained viable even after a prolonged treatment with high TNF concentrations. To evaluate whether a cause-effect relationship between A20-dependent NFkB inhibition and sensitization to TNF-induced death does exist, we comparatively analyzed a number of A20 mutants for their effect on NFkB activation and cellular viability following TNF treatment (Fig. 2a). Although the amino-terminal domain of A20 is capable of TRAF2 binding in a two hybrid system in yeast, it is not able to inhibit NFkB activation (Ref. 57 and Fig. 2b). Conversely, the carboxyl-terminal zinc-binding region (zinc domain), which does not inter-FIG. 1. Selective inhibition of TNF-induced NFkB activation by overexpressed A20. a, A20 does not block TNF-induced apoptosis in HeLa cells. HeLa cells were transiently transfected with a ␤-galactosidase expression vector (200 ng) together with the indicated expression vectors or empty vector (2 g) as indicated. 24 h after transfection cells, were treated with TNF (20 ng/ml) and CHX (10 g/ml) for 16 -20 h. After treatment, cells were fixed and stained with 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside as described. Intensely staining, shrunken blue cells showing loss of adherence were considered apoptotic. Data are expressed as the mean percentage (Ϯ S.E.) of apoptotic blue cells as a fraction of the total number of blue cells counted. Data are from six independent experiments (in each experiment at least 300 cells/experimental point were counted). A representative ␤gal assay is also shown. b, A20 effects on TNF-induced NFkB activity. For NFkB assays (25), HeLa cells were cotransfected with a NFkB-luciferase reporter plasmid (500 ng), together with the indicated expression vectors (1 g each), keeping the total amount of DNA constant with empty vector. Where indicated, cells were treated with TNF (20 ng/ml, 6 h) or phorbol myristate acetate (PMA) (100 ng/ml, 6 h). Data are from six pairs of experiments and are expressed in terms of luciferase activity relative to that of control cells.
FIG. 2. Sensitization to TNF-induced death by A20 correlates with NFkB inhibition. a, schematic representation of the A20 mutants used in the study. The schematic A20 structure with the seven zinc fingers is shown. Details on plasmids construction are given under "Materials and Methods." Expression of Flag-tagged A20 mutants in transfected HeLa cells is shown on the left. b, inhibition of TNF-induced NFkB activity by A20 mutants and correlation with sensitization to TNF-induced death. HeLa cells were cotransfected with NFkB-luc reporter (500 ng), pCDNA-HislacZ (200 ng), and the indicated expression vector (1.5 g). 24 h after transfection, half of the plates were treated with TNF (20 ng/ml) for 16 -20 h, and the remaining half were left untreated. NFkB assays and ␤gal assays were performed as described above. The results are representative of three independent experiments. act with TRAF2, is both sufficient and required for NFkB inhibition (Fig. 2b). Mutational analysis of the zinc domain showed that the last four zinc fingers (plasmid C4Zn) were both sufficient and required to provide nearly complete NFkB inhibition. Significantly, deletion of the last (seventh) zinc finger from either the zinc domain (plasmid Zndel1) or from C4Zn (plasmid C4Zndel1) severely compromised NFkB inhibitory activity, thus indicating a requirement for the last zinc finger for NFkB inhibition. Moreover, the seventh zinc finger is capable of inhibiting NFkB activity when oligomerized (Fig. 2b). The ability of each A20 mutant mentioned above to inhibit NFkB strongly correlated with its ability to sensitize HeLa cells to TNF-induced apoptosis, thus indicating that the A20 sensitizing effect is highly likely to be due to NFkB inhibition. As observed with full-length A20, a large fraction of the cells remained viable despite the strong NFkB inhibition.
NFkB Inhibition by IkB␣ Super-repressor Partially Sensitizes to TNF-induced Death-To determine whether NFkB inhibition tends to provide the same level of TNF sensitization irrespective of the blocking agent used, we studied the effects of an additional selective trans-dominant inhibitor of NFkB activation. The phosphorylation-defective IkB␣S32A/S36A acts by sequestering the cytoplasmic NFkB pool in a manner that is insensitive to extracellular stimuli (27,(35)(36)(37)(38)(39). IkB␣S32A/ S36A reduced both basal and induced NFkB activity even more strongly than A20 (Fig. 3); morphological analysis of TNFtreated transfected cells showed about 35% apoptotic cell death, i.e. an increase above the basal level that is comparable to (or slightly higher than) that obtained with A20-induced NFkB blockade. A similar degree of sensitization was obtained in Chang liver cells expressing IkB␣Ser32/36Ala (Fig. 3), as well in several other cell types. 3 This degree of sensitization was not significantly affected by various experimental parameters, such as the time lag between transfection and TNF treatment (between 12 and 48 h) or the length of TNF stimulation (between 12 and 24 h). Taken together, these results suggest the existence of cytoprotective signaling pathways that work despite a blocked NFkB-dependent transcription.
Sensitization to TNF Cytotoxicity and NFkB Blockade by Alternative Signaling-deficient TRAF2 Mutants-To examine whether a generalized TRAF2 inhibition provokes effects similar to those of selective NFkB inhibition, we tested a number of TRAF2 mutants interfering with TNF-R1 signaling by alternative mechanisms for their effect on NFkB activity and cell viability following TNF treatment. Deletion of the amino-terminal Ring finger (TRAF2 87-501) or of the most amino-terminal region (TRAF2 226 -501) generates signaling-deficient molecules that are capable of TRAF2 homodimerization and interaction with TRADD, thereby acting as dominant negative mutants (17,59). Despite a similar NFkB inhibition, both mutants sensitized to TNF-induced death much more effectively than IkB␣S32A/S36A (up to 85% apoptotic cells) in both HeLa and Chang cells (Fig. 4). Unlike IkB␣S32A/S36A, these mutants also block JNK/SAPK activation by TNF (27)(28)(29): therefore, their stronger sensitizing effect could be ascribed to concomitant JNK/SAPK inhibition. However, blockade of JNKs/ SAPKs by two different dominant negative JNK kinase (JNK kinase/SAP or ERK kinase) mutants (41, 42) did not alter HeLa or Chang cells sensitivity to TNF (Fig. 5a); more importantly, simultaneous NFkB and JNK/SAPK blockade by coexpression of IkB␣S32A/S36A and SEK-AL or SEK-KR did not generate potentiated biological response with respect to IkB␣S32A/S36A alone (Fig. 5a). Unlike SEK-AL, a dominant negative c-Jun mutant (c-Jun⌬169), which acts as a general inhibitor of AP1dependent transcription (43,60,61) without affecting NFkB activation, sensitized Chang but not HeLa cells to TNF-induced death (Fig. 5b), thus pointing to the existence of NFkB-independent and cell type-specific cytoprotective pathways. The different behavior of SEK-AL and c-Jun⌬169 is not surprising, as the latter is a much stronger inhibitor of AP1-dependent transcription than SEK-AL (25). 3 TRAF2(1-355) is a signaling-deficient mutant that retains the whole amino-terminal region and still interacts with TRAF2 but not with TRADD or NIK (26,59). We found that TRAF2(1-355), in addition to being unable to activate NFkB, is a powerful dominant negative inhibitor of NFkB activation by TNF (Fig. 4). This mutant also provided a degree of sensitization to TNF cytotoxicity that is comparable to that obtained with the amino-terminal TRAF2-deletion mutants described above.
TRAF6, a TRAF family member that has been implicated in interleukin 1-R1 signal transduction (62) is capable of TRAF2 interaction upon overexpression (62). 3 Consistent with its TRAF2-interacting ability, a dominant negative TRAF6 mutant lacking most amino-terminal region (TRAF6 275-522) partially interfered with TNF-induced NFkB activation in both HeLa and Chang cells (Fig. 4). The sensitizing effect of TRAF6⌬N to TNF cytotoxicity was lower than that of TRAF2(226 -501) but much higher than that of IkBS32A/S36A (Fig. 4): therefore, despite affecting NFkB activation by TNF at a lesser degree than IkBS32A/S36A, TRAF6⌬N was more powerful at sensitizing cells to TNF cytotoxicity, thus suggesting an interference with additional NFkB-independent cytoprotective functions of TRAF2.
Effects of NFkB-activating TRAFs on Cellular Sensitivity to TNF-induced Apoptosis-TNF pretreatment renders cells re- sistant to a subsequent challenge with TNF ϩ CHX, thus indicating that the induction of protective genes has occurred (Fig. 6a). We asked whether TRAF2 is sufficient to provide protection from TNF ϩ CHX. Despite activating NFkB more strongly than TNF, overexpressed TRAF2 did not provide protection from TNF ϩ CHX (Fig. 6a). TRAF6 overexpression activated NFkB to a similar extent and did not provide any degree of protection (Fig. 6a). TRAF2(1-358), a TRAF2 deletion mutant that, unlike TRAF2(1-355), retains the ability to bind NIK, activates NFkB much more strongly than TRAF2 or TRAF6 (26, 59) (Fig. 6a); however, unlike full-length TRAF2, this mutant is unable to bind TRADD (59). Similarly to TRAF2, and despite the prominent NFkB induction, this mutant is unable to protect cells from TNF ϩ CHX (Fig. 6a). Conversely, TRAF2(1-358) sensitized cells to TNF cytotoxicity to a maximal extent (Fig. 6b). The simplest interpretation of these results is that TRAF2-induced NFkB activation, although required for a complete cytoprotective response, is itself not sufficient to protect cells from TNF ϩ CHX. Therefore, a cell wherein NFkB has been hyperinduced by TRAF2 (1-358) is not  TNF-resistant: upon subsequent challenging with TNF, the TRAF2(1-358) mutant blocks endogenous TRAF2 recruitment to the TNF-R1/TRADD complex, the protective response does not take place, and the cell undergoes apoptosis. Consistent with the inability of NFkB to provide protection from TNFinduced cytotoxicity, neither interleukin 1 pretreatment nor p65/RelA transfection afforded protection from TNF ϩ CHXinduced apoptosis (Fig. 6a).
NFkB-independent Cytoprotection in Terminally Differentiated Primary Cells-The results shown above have been obtained in continuously cultured tumorous cell lines. To determine whether NFkB-independent cytoprotective pathways contribute to TNF resistance of terminally differentiated primary cells, we performed analogous experiments in primary rat oligodendrocytes, the glial cells responsible for the elaboration of the multilamellar myelin sheaths in central nervous system. Damage of oligodendrocytes in demyelinating diseases, such as multiple sclerosis, has been linked to TNF effects on these cells (63); however, primary rat oligodendrocytes are resistant to TNF cytotoxicity under standard culture conditions (46). We therefore infected purified primary rat oligodendro-cytes with recombinant retroviruses expressing either Flag-TRAF2(87-501) or HA-IkB␣S32A/S36A. After infection, cells were selected with puromycin for 4 days; the resulting polyclonal oligodendrocytes population was pretreated for 16 h with interferon ␥, which up-regulates TNF-R1 expression and renders oligodendrocytes responsive to TNF. 2 Cells were then TNF-stimulated for 6 h, fixed, and sequentially stained with anti-epitope antibodies and with terminal deoxynucleotidyltransferase-mediated dUTP nick end labeling to detect DNA fragmentation. TRAF2(87-501)-expressing oligodendrocytes were strongly sensitized to TNF-induced apoptosis (Fig. 7), thus indicating that also in these cells TRAF2 provides an essential contribution to cytoprotection. The sensitizing effect of TRAF2(87-501) was much greater than the one provided by retroviral transduction of IkBS32A/S36A (Fig. 7), thus indicating the existence of TRAF2-dependent protective pathways insensitive to NFkB blockade. DISCUSSION Tumor necrosis factor is capable of transducing cytotoxic signals mainly through p55TNF-R1, which is widely and con- Purified oligodendrocytes precursors were obtained from neonatal rat mixed glial primary cultures. Cells were incubated with growth factors (basic fibroblast growth factor and platelet-derived growth factor) and with the indicated recombinant retroviruses for 36 h. The medium was then replaced with defined serum-free differentiation medium. Retroviral infection did not interfere with terminal differentiation, as assessed by anti-galactocerebroside staining (data not shown). After differentiation, cells were pretreated with interferon ␥, which up-regulates TNF-R1 levels and renders oligodendrocytes responsive to TNF stimulation (46), 2 and then treated with mouse recombinant TNF for 6 h. Double staining for the retroviral expressed transgene and for DNA breaks (terminal deoxynucleotidyltransferase-mediated dUTP nick end labeling) was performed. From top to bottom: first row, Babe Puro; second row, IkBS32A/S36A; third row, TRAF2(87-501). Left, untreated cells; right, TNF-treated cells. Both virally transduced products were equally effective at inhibiting p65/RelA nuclear translocation in TNF-stimulated cells, as judged by anti-p65/RelA staining in parallel plates (data not shown). stitutively expressed (3). Apoptotic signaling occurs through sequential recruitment of the signal transducers TRADD and FADD, which in the end recruits and activates the apoptotic proteases (1,2). Despite being connected to the caspase enzyme cascade, the engagement of which directly triggers apoptosis, p55TNF-R1 engagement usually does not determine apoptotic cell death. The reason for this is that TNF induces a complex response that protects stimulated cells from apoptosis. The need for such a cytoprotective response arises from the strong inducibility of TNF synthesis and secretion in a very high number of pathological conditions (reviewed in Ref. 64); therefore, a vigorous and prevailing protective mechanisms is absolutely required to avoid unwanted death of cells exposed to TNF.
In most cell types, the cytoprotective response can be completely abrogated by protein or RNA synthesis inhibitors, such as CHX and ActD, thus indicating that transcription and translation of TNF-induced genes play a required role in cytoprotection. The molecular mechanisms of the cytoprotective response have been only partially elucidated. The signal transducer TRAF2, which is recruited to TNF-R1 via interaction with TRADD (5, 11), transmits signals required for activation of both JNK/SAPK and NFkB (11,17,(27)(28)(29); although the precise role of JNK/SAPK in regulation of programmed cell death is still a matter of debate, NFkB activation has been linked to transcriptional induction of protective genes by a number of experimental evidences. First, embryonic fibroblasts (EFs) from p65/RelA-deficient mice are extremely sensitive to TNF cytotoxicity (34); this result is intriguing because the other transcriptionally active NFkB/Rel subunits are supposed to be still functional in these cells. Therefore, it can be hypothesized that at least in EF cells, only a subset of NFkB-responsive genes specifically requiring p65/RelA to be transcribed carry out a protective function. The second evidence for a NFkB role in protection from TNF cytotoxicity arises from transfection studies performed with a dominant negative IkB␣ mutant bearing mutations in the amino-terminal amino acids (Ser 32 and Ser 36 ), the inducible phosphorylation of which, in response to TNF and other extracellular stimuli, targets IkB␣ to ubiquitindependent degradation (37)(38)(39). Both transient and stable expression of this IkB super-repressor (which nearly completely abrogates NFkB-dependent transcription) sensitizes various cell lines to TNF-induced apoptosis (35,37,64). In the cellular systems that we tested, we found that NFkB inhibition obtained by expression of either IkB super-repressor or the zinc finger protein A20, which shares with IkB super-repressor the ability to suppress TNF-induced NFkB activation in an apparently selective manner, significantly impaired the ability of cells to overcome TNF cytotoxicity. However, this sensitizing effect always affected only a fraction of transfected cells, and under no condition tested were we able to completely shut down the protective response. The fact that the cell types tested can be completely sensitized to TNF-induced apoptosis by blocking TNF-stimulated new gene expression (by use of the protein synthesis inhibitor CHX) but can only partially be sensitized by selectively shutting down NFkB-dependent transcription, suggests that NFkB-independent events occurring at the transcriptional/translational level are required to protect cells from TNF cytotoxicity. One practical implication of this is that it should be possible to identify TNF-induced protective genes in cells with a blocked NFkB-dependent transcription: this would hamper TNF induction of a huge number of NFkB-dependent genes (such as those encoding for cytokines and adhesion molecules) that are not relevant for cytoprotection.
Unlike a selective NFkB blockade, knocking out TRAF2 function by alternative strategies results in the complete abro-gation of the cytoprotective response. This indicates that in both continuously cultured tumor cells and highly differentiated primary cells (such as primary oligodendrocytes), NFkBindependent cytoprotective signals flow through TRAF2. Indeed, we have found evidence for AP1-dependent and NFkBindependent cytoprotective pathways in Chang but not HeLa cells, thus indicating that some of these cytoprotective responses may be specific for individual cell types. The previous observation that activation of transcription factor AP1 by p55TNF-R1 requires TRAF2 (25) is further evidence of the TRAF2-dependence of at least one of these additional cytoprotective pathways. Whether NFkB-independent TRAF2-elicited signals aimed at protecting cells from TNF cytotoxicity are exclusively transcriptional (i.e. devoted to transcriptional activation of protective genes), however, remains to be determined. Indeed, transcription and translation of protective genes may require some hours; thus, it is not at all clear why cells do not die before this protective protein synthesis has occurred. Activation of the apoptotic caspase pathway by a strictly related receptor, Fas/APO1/CD95, occurs in tens of seconds, with FADD and the first protease of the pathway (MACH/FLICE) almost instantly recruited to the engaged receptor (65). By analogy, we can hypothesize that to avoid the unwanted propagation of the death signal from TNF-R1 before completion of protective protein synthesis, the death pathway must be kept in a standby status (maybe by mechanisms interfering with FADD recruitment to TRADD or with activation of caspases). Although the experimental systems we used did not allow us to probe this hypothesis, the fascinating possibility that TRAF2 is also involved in this transcription-independent protective mechanism do exist.
The recent availability of TRAF2-deficient mice obtained by gene targeting has provided additional relevant information on the physiology of TRAF2, as well as on the mechanisms implicated in protection from TNF-induced apoptosis (66). Several TRAF2Ϫ/Ϫ cell lineages (including thymocytes and hematopoietic precursors) showed extreme sensitivity to TNF-induced cell death, thus indicating the abrogation of TRAF2-dependent cytoprotection. Extensive characterization of TRAF2 Ϫ/Ϫ EF cells provided some puzzling results. Although TRAF2 deficiency resulted in the absence of TNF-induced JNK/SAPK activation (consistent with the described TRAF2-dependence of this pathway) (27)(28)(29), NFkB activation was only delayed and partially reduced but not abolished. TRAF2-independent NFkB activation has also been described in transgenic mice expressing a dominant negative Traf2 mutant in a lymphocyte-specific manner (67). These results are quite surprising in the light of the previous studies addressing TRAF2 (17) and its interacting partner (and downstream effector) NIK (18,25,26) as the TNF-R1 signal transducers responsible for the direct activation of the IkB-kinase complex (20 -24) and consequently for NFkB activation. The involvement of TRAF2 in NFkB activation by TNF was originally suggested by transfection studies exploiting the signaling-deficient and dominant negative mutant TRAF2(87-501) (17): one possibility is that inhibition of TNFactivation of NFkB by this mutant reflects the titration and sequestering not of endogenous TRAF2 but of other TRAF2interacting signaling molecules involved in NFkB activation (such as RIP) (14,31). However, it is also possible that TNF-R1 is capable of activating NFkB through both TRAF2-dependent and TRAF2-independent pathways and that the relative contribution of the two pathways differs in various cell types. In support of this hypothesis, we have found that expression of dominant negative TRAF2 tends to affect NFkB activation by TNF in a cell type-dependent manner; however, most of the cell types tested were sensitive to its inhibitory effect. 3 As regards cellular sensitivity to TNF, TRAF2Ϫ/Ϫ EF cells differed from thymocytes and hematopoietic precursors in that they were TNF resistant; interestingly, however, TRAF2Ϫ/Ϫ cells (unlike wild type EF) were extremely sensitive to TNF in the presence of CHX, thus indicating that TRAF2 is required for a protein synthesis-independent cytoprotective signal.
Our results also indicate that, although required for TNFinduced cytoprotection and despite the prominent NFkB induction, TRAF2 is not sufficient to protect cells from a subsequent challenge with TNF ϩ CHX. This would suggest that additional TNF-R1-complex components provide a key contribution to the cytoprotective response. Consequently, NFkB activation through the TRAF2 pathway, although required for a complete cytoprotective response, is not itself sufficient to provide any kind of protection. The fact that NFkB activation is not sufficient to switch on the synthesis of protective genes makes sense; indeed, being responsive to a huge number of stimuli, NFkB is able to signal the occurrence of many different extracellular events. Consistent with this, the target genes of NFkB are numerous and are implicated in different cellular functions. Therefore, there must be mechanisms (such as the interaction of NFkB subunits with additional transcription factors that are activated in a stimulus-specific manner) allowing a restricted and specific program of gene expression to be activated in response to a particular inducing agent. Conversely, NFkB activation is usually required for the induction of NFkBcontaining promoters (58,68,69) (including those of putative protective genes), and this would explain the sensitizing effect of both IkB super-repressor and A20.
In this context, the effects of the TRAF2(1-358) mutant on cell viability are extremely interesting. This mutant is a stronger NFkB activator than full-length TRAF2; however, having its TrafC domain detected, it is unable to interact with TRADD (59). When overexpressed in transfected cells, TRAF2(1-358) is unable to provide protection from TNF ϩ CHX and conversely sensitizes cells to a subsequent challenge with TNF. Our interpretation of these results is that the strong NFkB activation provided by TRAF2 (1-358) is not sufficient to switch on the synthesis of cytoprotective genes; on the other hand, this mutant may either interfere with TNF-dependent Traf2 recruitment to TNF-R1/TRADD or titrate TRAF2-bound protective proteins (such as the inhibitors of apoptosis cIAP1 and cIAP2) (47), thus blocking the activation of the cytoprotective response. Irrespective of its mechanism of action, this mutant shows that TRAF2-dependent NFkB activation and cytoprotection can be dissociated and that a cell in which NFkB has been activated can remain exquisitely sensitive to TNF cytotoxicity if the TNF-R1/TRAF2 complex is inappropriately disturbed. Moreover, the observation that both NH 2 -terminal and COOHterminal TRAF2 deletion mutants are capable of sensitizing cells to TNF-induced apoptosis indicates that both deleted regions are required for activation and maintenance of the cytoprotective response.
In conclusion, the results reported in this paper indicate that TNF-inducible resistance to TNF cytotoxicity is a complex function and NFkB is only one component of the cytoprotective apparatus. Although cellular systems have been described in which NFkB blockade causes a maximal degree of TNF sensitivity (27, 34 -36), in most cell types we have tested so far, NFkB blockade usually did not provide a complete sensitization to TNF cytotoxicity, but always negatively affected cellular resistance to TNF in a variable percentage of cells. Conversely, interfering with TRAF2 function(s) nearly completely abrogated the cytoprotective response, thus suggesting that NFkBindependent protective genes are switched on upon TNF-R1 cross-linking in a TRAF2-dependent manner. The required but not sufficient role of TRAF2 in mounting the protective response points to a requirement for additional TNF-R1 complex components, and once more suggests that the integration of multiple signals must occur. We propose the existence of both generally acting (such as NFkB) and cell type-specific (such as AP1 in some cell lines, including Chang cells) cytoprotective components differentially contributing to the high threshold of TNF resistance of most cell types (Fig. 8). This highly integrated system would both provide multiple potential checkpoints and a high security level toward the unwanted elimination of cells exposed to TNF. FIG. 8. Protection from TNF cytotoxicity as a complex function depending on the integration of multiple signal transduction pathways. The figure shows TNF-R1-originating signal transduction pathways involved in protection from TNF cytotoxicity. TRAF2dependent and TRAF2-independent signals are further grouped into pathways that are sensitive or not sensitive to NFkB blockade.