NFκB Controls the Balance between Fas and Tumor Necrosis Factor Cell Death Pathways during T Cell Receptor-induced Apoptosis Via the Expression of Its Target Gene A20*

Activation-induced cell death (AICD), a term originally coined for the anti-CD3-induced apoptosis of T cell hybridomas and thymocytes, is predominantly driven by death receptors and has been involved in the control of autoreactive T cells in the periphery. In the Do-11.10 T cell hybridoma model of AICD, activation of the T cell receptor (TCR) results in Fas-dependent apoptosis. Here, we show that inhibition of the transcription factor nuclear factor κB (NFκB) in Do-11.10 cells resulted in increased sensitivity to TCR-mediated apoptosis, correlating with defective induction of the anti-apoptotic NFκB target gene A20. Stable expression of the zinc finger protein A20 in NFκB-negative Do-11.10 cells rescued the phenotype. TCR activation in NFκB-deficient Do-11.10 cells resulted predominantly in tumor necrosis factor (TNF) receptor 2 (TNFR2)-dependent bystander cell death rather than classical Fas-dependent AICD. Strikingly, A20 blocked TNF-mediated apoptosis and simultaneously restored TCR-induced Fas-dependent AICD. In addition, NFκB downstream of TNFR was required for up-regulation of Fas expression by endogenous TNF secreted in response to TCR stimulation. Together, these results suggest that NFκB can play both pro- and anti-apoptotic roles during AICD. We propose that NFκB controls the balance between Fas and TNF cell death pathways during AICD via the expression of the zinc finger protein A20.

T lymphocytes are highly specialized cells of the adaptive immune system. Individual T cells express distinct T cell receptors (TCRs) 1 able to recognize a particular antigen and initiate a properly directed immune response. The clonal deletion of auto-reactive T cells in thymus is essential for the discrimination between self and non-self (1,2). This "central tolerance" is, however, incomplete, and healthy individuals harbor self-reactive T cells in the periphery (3). Normally, such self-reactive T cells fail to induce autoimmunity as a result of "peripheral tolerance" mechanisms, which involve either induction of anergy or apoptosis in self-reactive T cells. Peripheral deletion by induction of apoptosis is a major mechanism of controlling self-reactive T cells, and two distinct molecular pathways are now recognized, which result in the induction of apoptosis in lymphocytes destined to die. Activation-induced cell death (AICD) is an antigen-dependent cell-autonomous process, which involves TCR-mediated sensitization to apoptosis followed by death receptor-dependent cell death (4). AICD is insensitive to action of anti-apoptotic Bcl-2 family proteins and does not rely on cytochrome c released from mitochondria. The second pathway of peripheral T cell deletion has recently been termed activated T cell autonomous cell death (5). In contrast to AICD, activated T cell autonomous cell death can be blocked by Bcl-2, is independent of death receptors, such as Fas, and is promoted by the pro-apoptotic Bcl-2 homologue Bim.
An important question in molecular immunology is the mechanism by which TCR induces apoptosis during AICD. T cell hybridomas represent a model system to study AICD in vitro (6). In this system TCR signaling triggers up-regulation of Fas ligand (FasL) at the cell surface, which leads to the engagement of Fas, followed by cell-autonomous cell death. Several transcription factors, such as c-Myc, NF-AT, Egr, and Nur77 (7)(8)(9)(10), have been shown to participate in the TCRmediated FasL up-regulation and induction of apoptosis in T cells. Nuclear factor B (NFB) transcription factors are also involved in the regulation of TCR-induced cell death (11)(12)(13) but the molecular mechanisms by which Rel/NFB proteins regulate TCR-induced apoptosis are still poorly understood.
NFB constitutes a family of evolutionary closely related transcription factors with important roles in immune and stress responses, inflammation, and cancer (14 -16). Five members have been identified in vertebrates: p50, p52, RelA (p65), RelB, and c-Rel. DNA binding activity of Rel/NFB complexes is regulated by members of the IB family, and several distinct IB molecules with conserved ankyrin repeats have been described. In most cell types, Rel/NFB proteins are trapped in the cytoplasm by the IB inhibitors. A wide range of stimuli activates the IB kinase complex, resulting in phosphorylation, ubiquitination, and degradation of IBs. Consequently, Rel/ NFB proteins translocate to the nucleus and bind to so-called B sequence motifs in regulatory regions of target genes (17,18). Abundant evidence has implicated Rel/NFB transcription factors also in the control of apoptosis (19,20). It is well established that NFB is required for protection from tumor necrosis factor (TNF)-mediated cell death in mouse embryonic fibro-blast, T cells, and many other cell types (20 -22). NFB activation in response to TNF receptor (TNFR) signaling, for instance, results in the induction of diverse anti-apoptotic proteins, which interfere at different levels with the apoptotic signal transmitted through the TNFR (20). With respect to TCR-mediated apoptosis, most reports document an anti-apoptotic role for NFB, but in some experimental systems NFB can also promote TCR-induced cell death (12,13).
Here, we show that in a widely used T cell hybridoma model of AICD NFB function downstream of TCR is largely limited to the control of the anti-apoptotic zinc finger protein A20. In the absence of NFB, hybridoma cells predominantly die in a cell non-autonomous TNF-dependent and Fas-independent fashion and overexpression of A20 completely reversed this phenotype. We propose that during AICD NFB controls the balance between Fas and TNF cell death pathways via the expression of A20, enabling cells to maintain clonal specificity of AICD.

EXPERIMENTAL PROCEDURES
Cell Culture, Induction, and Flow Cytometric Analysis of AICD-The murine T cell hybridoma Do-11.10 was obtained from the ATCC. Cells were cultured in Dulbecco's modified Eagle's medium supplemented with 5% heat-inactivated FCS (BioWhittaker), penicillin (100 units/ml), streptomycin (100 g/ml), and L-glutamine (2 mM) at 5% CO 2 , 37°C. PMA (10 nM) and ionomycin (0.5 M) were dissolved in Me 2 SO. Recombinant murine TNF (20 ng/ml) was from PromoCell. For induction of apoptosis, 24-well plates were precoated with 250 l of anti-CD3 monoclonal antibody (mAb) (clone 145-2C11, Becton Dickinson) diluted to 0.5 g/ml (unless otherwise stated) in 50 mM Tris-HCl, pH 9, at 4°C overnight. Plates were washed once with phosphate-buffered saline, and 1 ϫ 10 5 cells in 0.4 ml of medium were applied per well and incubated overnight (unless otherwise stated). Alternatively, cells were treated with agonistic anti-Fas mAb (clone Jo2, 1 g/ml, Becton Dickinson). After incubation cells were resuspended in fluorescence-activated cell sorting buffer (phosphate-buffered saline, 2% FCS) containing 2 g/ml propidium iodide. Cell death was measured by determining the fraction of propidium iodide-positive cells with a FACScan flow cytometer and CellQuest Pro software (Becton Dickinson). Similar results were obtained when cells were stained for annexin V (data not shown).
Antibodies and Blocking Reagents-For blocking experiments the following concentration of neutralizing mAbs (Becton Dickinson) were added to the wells at the beginning of the incubation: anti-TNFR1 (clone 55R-170, 2 g/ml), anti-TNFR2 (clone TR75-54, 2 g/ml), anti-FasL (clone MFL-3, 1 g/ml), and anti-TNF (clone TN3-19.12, 30 g/ ml). Recombinant human TNFR1-Fc fusion protein (Alexis) was used at 5 g/ml. Biotinylated anti-T7-tag mAb was purchased from Novagen; anti-FLAG mAb (clone M2) was obtained from Sigma; anti-IB␣, anti-RelA, and anti-RelB rabbit polyclonal Abs are described elsewhere (23). Streptavidin-HRP, goat anti-rabbit Ig-HRP, and goat anti-mouse Ig-HRP secondary reagents were from Dako. To generate rabbit polyclonal anti-mouse A20 Abs, the C-terminal portion of A20 (amino acids 560 -775) was amplified by PCR with recognition sites for NheI and XhoI incorporated into the 5Ј and 3Ј primers, respectively. The PCR product was subcloned into NheI/XhoI sites of pET28a (Novagen), which resulted in an in-frame fusion of the A20 coding region to His 6 tag sequences. His-tagged C-terminal A20 domain was expressed in BL21 (DE3) bacteria after induction with 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 5 h. Recombinant protein was purified on nickel-nitrilotriacetic acid resin (Qiagen) under denaturing conditions (8 M urea) according to recommendations from the manufacturer. Protein was dialyzed against 20 mM Tris-HCl, pH 8, 150 mM NaCl, 2 M urea and used for rabbit immunizations (Moravian Biotechnology).
Plasmid Constructs, Stable Transfection, and Retroviral Transduction of Do-11.10 Cells-A cDNA encoding non-degradable mIB␣ carrying alanines instead serines at positions 32 and 36 was subcloned into the HindIII site of pRcCMV (Invitrogen) via blunt end cloning. Do-11.10 cells were electroporated (5 ϫ 10 6 cells in 350 l of Dulbecco's modified Eagle's medium at 250 V, 960 microfarads, 0.4-cm cuvettes in a Bio-Rad Gene Pulser) and plated in 20 ml of fresh medium. After 48 h cells were counted and plated in 96-well plates at limiting dilution in medium containing 1.5 mg/ml Geneticin (Invitrogen). Single clones were picked, expanded, and tested for NFB induction in response to treatment with anti-CD3 mAb. Flag-tagged A20 cDNA (24) (kindly provided by Karen Heyninck and Rudi Beyaert) was inserted into the EcoRI site of pczCFG5 IEGZ (25) (kindly provided by Dirk Lindemann) via blunt end cloning. The Phoenix packaging cell line (26) (kindly provided by Andrea Denk and Thomas Wirth) was transiently transfected with pczCFG5 IEGZ or pczCFG5 IEGZ-A20 using LipofectAMINE 2000 (Invitrogen). After 2 days supernatants were collected and filtered through a 0.45-m filter. Two million exponentially growing Do-11.10 cells were resuspended in 2 ml of virus-containing supernatant. After addition of Polybrene (Sigma, 5 g/ml) cells were plated in 12-well plates and centrifuged for 3 h at 1000 ϫ g at room temperature. Cells were replated in 20 ml of fresh medium and after 48 h replated in 96-well plates at limiting dilution in medium containing 300 g/ml zeocin (Invitrogen). Single clones were picked and screened for green fluorescent protein expression by flow cytometry. Green fluorescent protein-positive clones were analyzed for A20 expression by Western blotting or for resistance to TNF-induced apoptosis. Pools of A20-overexpressing clones were used for further experiments.
Semi-quantitative RT-PCR Analysis of Steady State mRNA Levels-Preparation of total RNA and semi-quantitative RT-PCR analysis were performed as previously described (27). Primer sequences and amplification conditions are available upon request.
Mice and Generation of T Cell Blasts-Double-knockout tnf Ϫ/Ϫ lta Ϫ/Ϫ mice (28) were kindly provided by Dr. Heinrich Körner. C57BL/6 mice were used for splenocyte preparations. All animals were housed and bred under standardized conditions with water and food ad libitum in the SPF mouse facility of the Forschungszentrum Karlsruhe, Institute of Toxicology and Genetics. Total splenocyte suspensions were prepared according to standard procedures and plated in RPMI, 10% FCS, Lglutamine (2 mM), penicillin (100 units/ml), streptomycin (100 g/ml), 50 M ␤-mercaptoethanol at a density of 2 ϫ 10 6 cells/ml. Five ml of cell suspension were activated in anti-CD3 mAb precoated 6-well plates (5 g/ml) for 48 h. Viable blasts were counted and plated at a density of 2 ϫ 10 5 cells/ml in medium supplemented with IL-2 (PromoCell, 10 ng/ml), and cells were expanded for an additional 48 -72 h before being used for experiments.

Inhibition of NFB by a Non-degradable Mutant mIB␣ Results in Increased Sensitivity to TCR-induced Apoptosis-To
investigate the role of NFB in the T cell hybridoma model of AICD, we blocked NFB nuclear translocation by the stable expression of a non-degradable mutant mIB␣ molecule (29) in Do-11.10 cells. Several clones were isolated and two, termed Do/B pos (1) and Do/B pos(2) , did not show any expression of the mIB␣ transgene and were used as controls in all subsequent experiments. Two other clones, termed Do/B neg (1) and Do/ B neg (2) , showed strong mIB␣ expression with a concomitant decrease in endogenous IB␣ levels (Fig. 1A). The efficiency of NFB inhibition was analyzed in EMSAs with nuclear extracts from anti-CD3-activated cells. Whereas NFB was strongly induced in control cells (complex I), mIB␣-expressing clones showed no NFB binding with the exception of p50-p50 homodimers (complex II, Fig. 1B). In control cells, the majority of NFB binding was composed of p50-RelA heterodimers. At later time points (2-8 h), additional complexes containing c-Rel (complex III) and RelB (complex IV) were apparent in control but not in Do/B neg cells (Fig. 1B).
To demonstrate lack of NFB transcriptional activity downstream of TCR, Do-11.10 cells were stimulated with anti-CD3 mAb and induction of the classical NFB target gene nfkb2 was determined by RT-PCR. Expression of nfkb2 mRNA was induced in control cells with kinetics matching NFB binding as detected in EMSAs. In contrast, induction of nfkb2 was almost completely abolished in NFB-negative clones (Fig. 1C). To analyze the role of NFB in the regulation of AICD, TCRinduced cell death was compared in Do/B pos and Do/B neg cells. Anti-CD3 stimulation resulted in increased sensitivity to AICD in cells lacking NFB. Moreover, in control cultures marked apoptosis was detected only after 10 -12 h, whereas cell death occurred ϳ2 h earlier in Do/B neg cells (Fig. 1D). Thus, overexpression of the mIB␣ transgene in Do-11.10 hybridoma cells results in a complete block of NFB activation downstream of the TCR and in increased sensitivity to TCR-induced apoptosis.
TCR Signaling Induces NFB-dependent Expression of the Anti-apoptotic Zinc Finger Protein A20 -NFB is a well documented principal regulator of genes protecting cells from TNFmediated apoptosis (19,20). We therefore analyzed whether the increased TCR-induced apoptosis in NFB-deficient Do-11.10 cells correlated with impaired expression of NFB target genes. Pools of Do/B pos and Do/B neg clones were activated with anti-CD3 mAb or PMA/ionomycin and mRNA levels of 11 reported anti-apoptotic NFB target genes were determined by semi-quantitative RT-PCR. Expression of the egr3 gene, which is induced in an NFB-independent manner, was measured as a control for efficient TCR signaling. Table I shows the genes analyzed and their dependence on NFB. Interestingly, only c-IAP1 and A20 were both TCR-inducible and completely dependent on NFB activation, whereas a partial dependence on NFB was observed for the PMA/ionomycin-induced expression of the genes encoding Bfl-1/A1 and Bcl-X L (Table I and Fig. 2A). In addition, induction of egr3 gene expression was normal in all anti-CD3-treated samples, demonstrating specificity of the mIB␣ inhibitor ( Fig. 2A).
Because c-IAP1 has recently been shown to perform also pro-apoptotic functions (30), we focused on A20 as a genuine anti-apoptotic gene. Moreover, the fact that A20 expression was induced by TCR signaling was a surprising novel finding. Analysis of TCR-induced A20 mRNA levels in individual clones confirmed the impaired induction in pooled Do/B neg cells, and a quantification of this experiment is shown in Fig. 2B. To demonstrate that A20 mRNA induction also results in increased A20 protein levels, we raised Abs directed against the C-terminal portion of mouse A20. Western blot analysis dem- onstrated that A20 protein levels were clearly induced after 4 -8 h of anti-CD3 stimulation in control but not in NFBnegative Do-11.10 cells (Fig. 2C).
To address whether TCR stimulation also induces A20 expression in primary T lymphocytes, we generated T cell blasts from mouse splenocytes. Fig. 2D shows that TCR signaling strongly induced A20 expression on both mRNA and protein level. Similar results were obtained with purified CD4 ϩ and CD8 ϩ T cell blasts (data not shown). To clearly demonstrate that TCR signaling also induces A20 expression independent of TNF, we analyzed T cell blasts from tnf Ϫ/Ϫ lta Ϫ/Ϫ double-knockout mice, which are deficient in any known TNFR ligand (28). Anti-CD3-induced A20 levels were comparable between wild-type and tnf Ϫ/Ϫ lta Ϫ/Ϫ T cell blasts, indicating that activation of the TCR induces A20 independent of TNFR signaling (Fig. 2E). Together, these data indicate that TCR-induced A20 expression is a general feature of mature T cells and that this response is, at least in Do-11.10 cells, under NFB control.
Overexpression of A20 in NFB-negative Do-11.10 Hybridoma Cells Rescues Both Enhanced AICD and Sensitivity to TNF-To test whether A20 overexpression rescues Do/B neg cells from increased sensitivity to AICD, we transduced both Do/B pos and Do/B neg clones with either empty vector or a full-length A20 cDNA. Consistent with data from other cell systems (20,21), we found that Do/B neg cells were very sensitive to TNF-induced apoptosis and A20 completely rescued this phenotype (Fig. 3A). This result demonstrates that A20 levels in transduced cells were sufficient for effective protection from TNF. We also measured the influence of A20 on AICD. Overexpression of A20 had no apparent effect on AICD in Do/B pos cells. In Do/B neg cells, however, reconstitution with A20 reduced TCR-induced apoptosis to levels similar to the ones seen in control cells (Fig. 3B). A20 overexpression did not Ϫ ND a Indicates either unaltered (Ϫ) or increased (ϩ) mRNA levels upon anti-CD3 induction. b Dependence on NFB was determined by comparing Do/B pos and Do/B neg clones. c XIAP mRNA was not detected (mouse brain cDNA served as a positive control). affect dexamethasone-induced apoptosis, indicating that apoptosis is not generally blocked in this system (data not shown). We conclude that expression of A20 in Do/B neg cells is sufficient to rescue both TNF sensitivity and enhanced AICD in response to TCR signaling.
To determine A20 protein levels at the onset of AICD in reconstituted Do/B pos and Do/B neg clones, Western blots were performed. The amount of A20 protein in control cells was much lower than in cells transduced with an A20 cDNA. Moreover, anti-CD3 treatment increased steady state levels of transgenic A20, even in the absence of NFB. We also detected increased levels of endogenous A20 protein in anti-CD3-treated Do/B pos , but not in Do/B neg , clones (Fig. 3C). Pulse-chase experiments revealed that TCR signaling increased A20 protein synthesis ϳ2.5-fold, whereas the half-life of A20 (ϳ3.5 h) was not affected (data not shown).
Because it has been reported that high levels of A20 suppress NFB (31), we analyzed anti-CD3-mediated activation of NFB in Do/B pos cells overexpressing A20. Activation of NFB in A20-transduced Do/B pos cells was clearly impaired (see Supplemental Fig. 1, available in the on-line version of this article). However, the cells were not NFB-negative but instead showed a delayed activation, with NFB DNA-binding levels being ϳ33% reduced at 2-and 4-h time points compared with vectortransduced controls.
Inhibition of NFB in Do-11.10 Cells Results in a Switch from FasL-to TNF-dependent AICD, Which Is Reversed by A20 -In the Do-11.10 T cell hybridoma model of AICD, TCR signaling results in Fas/CD95-dependent apoptosis. Because Do/B neg cells showed increased sensitivity to TCR-induced cell death with apoptotic cells occurring ϳ2 h earlier than in controls, we asked whether cell death proceeds by a different molecular mechanism in Do/B neg compared with Do/B pos cells. Analyzing the involvement of different death receptors from the TNFR superfamily with specific blocking reagents confirmed FasL as principal inducer of cell death in anti-CD3treated Do-11.10 cultures. Unexpectedly, anti-TNF reagents also consistently reduced the level of AICD in Do-11.10 cell by ϳ50% (Fig. 4A). Reagents blocking TRAIL and TWEAK were ineffective (data not shown). In striking contrast to Do/B pos cells, AICD in Do/B neg cells was minimally reduced by anti-FasL mAb but very effectively blocked by anti-TNF reagents (Fig. 4A). Do/B pos cells overexpressing A20 showed AICD very similar to cells transduced with empty vector. Interestingly, Do/B neg cells reconstituted with A20 also showed normal AICD hardly distinguishable from Do/B pos controls (Fig. 4B).
These results indicate that, in the absence of NFB, the FasL/ Fas pathway operates inefficiently and instead cells die by endogenous TNF. Moreover, the reliance on the FasL/Fas pathway can be restored by A20 overexpression in Do/B neg cells, demonstrating that A20 is a major target of NFB downstream of the TCR.
The involvement of TNF in AICD of Do-11.10 cells was unexpected because A1.1 hybridomas were reported to exclusively rely on the Fas and not the TNFR pathway (32). Other studies, however, suggest that TNF sensitizes hybridoma cells to Fas-mediated apoptosis (33). We therefore stimulated Do-11.10 cells with TNF and agonistic anti-Fas mAb. Neither of these reagents alone was capable of inducing apoptosis (3-4% cell death after overnight treatment), whereas the combined stimulation resulted in extensive cell death (35%). Thus, TNF is capable of stimulating the Fas pathway, offering an explanation for the observed influence of endogenous TNF on AICD in Do-11.10 hybridoma cells.
Do-11.10 cells express both TNFR1 and TNFR2 (data not shown). To examine which TNFR is involved, Do-11.10 cells were stimulated with anti-CD3 mAb in the absence and presence of anti-TNFR1/R2 blocking mAbs. In Do/B pos , as well as in Do/B neg cells, blocking of TNFR2 had a much more pronounced effect, implying that TNFR2 mediates the majority of TNF signaling (Fig. 4C). The predominant involvement of TNFR2 in our AICD model may be the result of the fact that TNFR2 is preferentially triggered under conditions of limiting amounts of TNF (34). putative NFB target genes (12,35,36). Blocking of NFB did not affect anti-CD3-induced FasL mRNA levels or FasL surface expression (Fig. 5, A and B). Basal Fas surface expression was also unchanged, whereas anti-CD3-induced up-regulation of Fas mRNA and Fas surface expression was abolished in Do/ B neg compared with Do/B pos cells (Fig. 5, A and B).
Because endogenous TNF clearly contributed to the regulation of AICD in our system and TNF induces Fas in fibroblasts (37), we analyzed the involvement of endogenous TNF in Fas regulation by employing anti-TNFR1 or anti-TNFR2 neutralizing mAbs. Blocking TNFR1 and TNFR2 during AICD of Do/ B pos cells led to greatly reduced induction of Fas mRNA (Fig.  5B) and a complete block of anti-CD3-induced Fas up-regulation on the cell surface (Fig. 5C). Both TNFRs participated with slightly stronger involvement of TNFR2. Thus, anti-CD3-induced up-regulation of Fas in Do-11.10 cells is mediated by secreted TNF and dependent on NFB.
We further wanted to know whether the observed lack of Fas induction was responsible for the reduced Fas-mediated apoptosis in Do/B neg cells. We took advantage of the observation that Do/B pos cells and Do/B neg cells reconstituted with A20 showed comparable Fas-dependent AICD (see Fig. 3B). Fig. 5D demonstrates that A20 did not rescue the impaired anti-CD3induced up-regulation of Fas in Do/B neg cells despite apparently normal AICD, indicating that basal Fas expression is sufficient for normal AICD in Do-11.10 cells.
NFB Downstream of Both TCR and TNFR Is Required for Maximal A20 mRNA Induction-To examine how much of the anti-CD3-induced NFB activity was dependent on endogenous TNF, Do-11.10 cells were stimulated with anti-CD3 mAb in the absence and presence of anti-TNFR1/R2 neutralizing mAbs and nuclear extracts were analyzed in EMSAs. Blocking of both TNFRs revealed that the majority of NFB induction after 2 h of anti-CD3 treatment was mediated by endogenous TNF, although there was also TNF-independent NFB induction, which was even more pronounced after 4 and 8 h (Fig. 6A). Anti-TNFR1 mAbs did not significantly affect the NFB induction profile compared with control cultures. Blocking of TNFR2, however, resulted in a clearly reduced NFB induction, particularly after 4 and 8 h of stimulation (Fig. 6A).
A20 is a classical TNF-responsive gene (38). To determine the contribution of endogenous TNF to the induction of A20 mRNA, control Do-11.10 cells were stimulated in the absence or presence of anti-TNFR blocking mAbs. Both anti-TNFR mAbs lowered anti-CD3-induced A20 mRNA levels. In the presence of blocking anti-TNFR2 mAb, the effect of anti-TNFR1 mAb was hardly visible, again arguing for a major involvement of TNFR2 in this system (Fig. 6B). Together, these results demonstrate that during AICD NFB is induced by both TCRand TNF-dependent pathways, uncovering a previously unanticipated complexity of signals that regulate AICD. DISCUSSION The role of NFB in apoptosis has been studied in many different systems, but the molecular mechanisms by which NFB-regulated gene expression influences TCR-mediated cell death are still largely unknown. We have analyzed 11 previously described anti-apoptotic NFB target genes in the Do-11.10 T cell hybridoma model of AICD. Comparing NFBpositive and NFB-deficient Do-11.10 clones, we found that TCR signaling only regulates A20 and c-IAP1 genes in an NFB-dependent manner whereas the other NFB targets are expression plasmid. C, the individual contribution of TNFR1 and TNFR2 to anti-CD3-induced cell death in Do/B pos and Do/B neg clones was analyzed with mAbs specifically blocking TNFR1 or TNFR2. Quantification was performed as in Fig. 3 (A and B). 4. Inhibition of A20 induction by blocking NFB in Do-11.10 cells results in a switch from FasL-to TNF-dependent AICD. Pools of Do/B pos and Do/B neg clones were stimulated with anti-CD3 mAb for 16 h in the absence or presence of indicated death receptor blocking reagents. Apoptotic cells were detected by flow cytometry, and data are shown as apoptotic index, i.e. cell death in cultures treated with anti-CD3 mAb alone (Ϫ) was set to 100% and apoptosis that remained in the presence of indicated death receptor blocking reagents was plotted as percentage of total TCR-mediated cell death to demonstrate the relative contribution of Fas and TNFR signaling pathways. Absolute levels of apoptosis were similar to Fig. 3. A, Do/B pos and Do/B neg cells stably transduced with empty vector as control. B, Do/B pos and Do/B neg cells reconstituted with stably integrated A20 either not induced by TCR signaling or induced normally in the absence of NFB. It is important to note that A20 mRNA expression is also up-regulated by TCR signaling in primary T cell blasts (this report), but not in thymocytes or naïve mature T cells (39) (data not shown). Interestingly, anti-CD3-induced expression of Bfl-1/A1 and Bcl-X L is not dependent on NFB and induction by PMA ϩ ionomycin is only partially blocked in cells lacking NFB. These results show that many so-called NFB target genes are only facultatively dependent on NFB and that dependence on NFB has to be considered in the context of stimulus, target gene, and differentiation state, e.g. thymocyte versus T cell blast.
In contrast to its widely accepted anti-apoptotic function, it has been proposed that NFB can also play a positive role in TCR-induced death of T cell hybridomas (12) or thymocytes (13). In another report, no difference in TCR-mediated apoptosis was observed in T cells from nfkb1 Ϫ/Ϫ , relA Ϫ/Ϫ , or c-rel Ϫ/Ϫ mice (37). Because TCR signaling induces multiple Rel/NFB proteins (40), the lack of a single Rel/NFB family member may not result in a severe phenotype. On the other hand, naïve T cells from c-rel Ϫ/Ϫ relA Ϫ/Ϫ double-knockout and mIB␣-transgenic mice show greatly attenuated activationinduced proliferation, making it difficult to analyze AICD (37) (data not shown). One group that succeeded in measuring AICD in T cell blasts from mIB␣-transgenic mice also found increased apoptosis in response to TCR signaling (11). Our data show that in Do-11.10 cells inhibition of NFB results in increased sensitivity to TCR-mediated apoptosis. This is, however, a net effect because dissection of the mechanism of apoptosis revealed both increased TNF-induced apoptosis and markedly reduced FasL-dependent cell death in NFB-deficient Do-11.10 cells. Thus, in contrast to either anti-or proapoptotic roles for NFB, these results argue for both anti-and pro-apoptotic functions of NFB downstream of the TCR.
A20 Replaces NFB Function during AICD of T Cell Hybridomas-NFB deficiency in hybridoma cells results in de- creased ability to undergo FasL-dependent AICD and acquired sensitivity to die in response to secreted TNF. Because overexpression of the zinc finger protein A20 completely reverses this phenotype, it is possible that A20 also performs a dual function; it inhibits TNF-mediated apoptosis and at the same time restores TCR-induced FasL-dependent apoptosis. Recently, it has been demonstrated that A20 inhibits TNF-induced apoptosis by disrupting recruitment of TRADD and RIP to the TNFR1 complex in Jurkat T cells (41). In NFB-deficient Do-11.10 cells, TNFR2 is the receptor primarily inducing apoptosis and A20 completely inhibits this cell death. Because RIP was also reported to mediate TNFR2-induced apoptosis in T cells (42), it is likely that A20 blocks TNF-induced apoptosis in Do-11.10 cells by inhibiting recruitment of RIP to both TNFR1 and TNFR2.
The finding that A20 restores Fas-mediated AICD in Do-11.10 cells is surprising. Unfortunately, it is difficult to directly assess the action of A20 on Fas-mediated AICD because activated T cells secrete TNF and A20 is required to protect them from TNF-induced cell death. We did not observe any effect of A20 on the expression of both FasL and Fas, suggesting that A20 regulates this cell death pathway downstream of Fas. The dual role of A20 in TCR-induced apoptosis is important in the light of recent findings, indicating that some of the classical anti-apoptotic NFB target genes can also promote cell death. For example, c-IAP1 has been found to stimulate TNFR2-dependent degradation of the anti-apoptotic TRAF2 protein (30). In addition, c-FLIP has been reported to promote apoptosis by enhancing caspase-8 processing after Fas engagement. Interestingly, the pro-apoptotic function of c-FLIP operates at low but physiologically relevant levels whereas high ectopic expression levels result in rescue from apoptosis (43). Thus, the propensity of a given gene, and consequently NFB as its regulator, to antagonize apoptosis may be a context-dependent phenomenon rather than an invariant feature. A20 does not bind directly to death domain signaling molecules or death receptors, but it can be recruited to the TNFR1 complex via its interaction with TRAF2 and/or IB kinase ␥/NEMO (31,41,44). Several other proteins interact with A20, including ABIN-1 (24), ABIN-2 (45), and TXBP151 (46). As the biochemical function of A20 and these interacting proteins is still poorly understood, a molecular understanding of the precise role of A20 in TCR-induced apoptosis requires further investigation.
In agreement with the reported inhibitory function of A20 on NFB activation (31), we also observed reduced induction of NFB binding in anti-CD3-stimulated Do-11.10 cells overexpressing A20. This result does not change our interpretation with respect to AICD because even a complete block of NFB by a non-degradable mutant mIB␣ can be rescued by A20 overexpression, indicating that as long as A20 is induced by NFB downstream of the TCR (or provided constitutively) AICD proceeds normally.
Role of TNF and TNFR2 in Fas-dependent Cell Death during AICD of Do-11.10 Cells-The observation that endogenous TNF is necessary for optimal AICD of Do-11.10 cells is unexpected because an exclusive requirement for the FasL/Fas system was reported for the A1.1 hybridoma model of AICD (32). One possible explanation is that different hybridomas show different sensitivity toward Fas-induced apoptosis. In accordance with our data, it was shown that effective Fas-mediated apoptosis of Do-11.10 cells and primary CD8 ϩ T cell blasts requires pre-treatment with TNF and functional TNFR2, respectively (33,47). In addition, stimulation of naïve T cells with  6. TCR-stimulated secretion of TNF contributes to both maximal NFB binding and A20 mRNA induction. A, nuclear extracts from anti-CD3-stimulated Do-11.10 cells were prepared and analyzed in EMSAs as described in Fig. 1B with the exception that anti-TNFR neutralizing mAbs were included during the stimulation period as indicated. B, A20 mRNA levels were analyzed by RT-PCR as described in Fig. 5B. ␤-actin levels are shown as amplification control. un, untreated culture.
anti-TNFR2 agonistic mAb results in enhanced sensitivity to Fas cross-linking (33). Taken together, these results suggest a general positive influence of TNF and TNFR2 on the Fas pathway. Indeed, there is growing evidence indicating that TNFR2 is the major TNF receptor in T cells, in particular for the induction of apoptosis (42,48). This raises the possibility that other TNFR superfamily members, which also lack a death domain and therefore have been excluded as potential apoptosis inducers, may turn out to participate in the regulation of apoptosis.
With the help of relA Ϫ/Ϫ mouse fibroblasts, it has been shown that NFB is essential for TNF-induced up-regulation of Fas (36,37). We demonstrate that a similar requirement also exists in T cells, suggesting that Fas up-regulation during AICD could sensitize to apoptosis by increasing the probability of Fas engagement on the cell surface. This was, however, ruled out by the finding that Fas basal levels on T cell hybridoma are sufficient for effective AICD, indicating that TNF-mediated sensitization to AICD occurs downstream of Fas.
TCR Activates A20 Gene Expression: Differential Regulation at mRNA and Protein Levels-NFB-dependent induction of A20 mRNA expression downstream of both TCR and TNF is consistent with earlier reports demonstrating functional NFB-responsive elements in the A20 promoter (49). The regulation of A20 protein levels by the TCR, however, seems unusual. A20 mRNA levels are maximal after 2 h of TCR signaling, whereas A20 protein is not significantly induced at this time point, reaching maximal levels only after 4 h. Steady state A20 protein, but not mRNA, levels expressed from a retroviral vector also increase upon TCR signaling. This effect is not NFB-dependent and correlates with increased A20 synthesis as indicated by pulse-chase experiments (data not shown). Thus, A20 gene expression is controlled at both transcriptional and translational level. A similar mode of regulation has been described for another anti-apoptotic NFB target gene, XIAP (50), indicating that multi-level regulation of genes controlling apoptosis is more common than previously appreciated.
Role of NFB in AICD-Our data suggest a model in which NFB activated by the TCR and subsequently by secreted TNF induces production of A20. As a consequence, A20 inhibits TNF-induced cell death and at the same time supports FasLdependent apoptosis, performing both anti-and pro-apoptotic roles in AICD (Fig. 7). AICD is involved in both removal of activated T cells after an immune response and in balanced expansion of T cells during the activation phase (6). This is critical because the T cell compartment has limited capacity and a particular T cell clone should not dominate the immune response. Therefore, AICD operates in strictly antigen-dependent manner so that only TCR-stimulated T cells undergo apoptosis. In the absence of NFB, however, TNF secreted by TCR-stimulated cells kills NFB-deficient cells in trans. This so-called bystander cell death, in which also non-activated T cells die, is highly undesirable because it disrupts the physiological function of AICD (Fig. 7). Thus, NFB via the expression of its target gene A20 contributes to the maintenance of clonal specificity during AICD.