Apoptotic crosstalk of TNF receptors: TNF-R2-induces depletion of TRAF2 and IAP proteins and accelerates TNF-R1-dependent activation of caspase-8

Tumor necrosis factor (TNF) exerts its biological functions by interactions with two members of the TNF receptor superfamily, namely TNF-R1 and TNF-R2. The cytoplasmic tail of TNF-R1 contains a death domain (DD), which is essential for induction of apoptosis. However, this motif is missing in TNF-R2 and the function of the latter receptor is poorly understood. TNF occurs in two bioactive forms. In its membrane-bound form (mTNF) TNF stimulates TNF-R1 and TNF-R2, whereas in its proteolytically processed soluble form (sTNF) TNF predominantly signals via TNF-R1 (Grell et al., 1995; Grell et al.,1998). A variety of reports indicate that TNF-R1 triggers the apoptotic program by sTNF-induced sequential recruitment of the adaptor protein TRADD, the TRADD-interacting adaptor protein FADD and the FADD-interacting initiator caspase FLICE/caspase-8(Locksley et al., 2001;Wallach et al., 1999). However, activation of TNF-R1 also results in a strong activation of NF-κB and induction of NF-κB-regulated anti-apoptotic factors by a pathway including TRADD, TRAF2, RIP (for a review, seeWajant et al., 1999;Wajant et al., 2001) and MEKK3(Yang et al., 2001). Consequently, TNF-R1-induced apoptosis typically develops only when the anti-apoptotic NF-κB response is downregulated. For example, mice deficient in components of the TNF-induced NF-κB response (RIP, TRAF2,IKKβ, IKKγ, p65, MEKK3) die because of massive liver apoptosis, and embryonal fibroblasts derived from these mice show enhanced sensitivity to TNF-induced apoptosis (Beg et al.,1995; Kelliher et al.,1998; Li et al.,1999; Makris et al.,2000; Rudolph et al.,2000; Schmidt-Supprian et al.,2000; Tada et al.,2001; Tanaka et al.,1999).

After ambiguous results concerning the apoptosis-inducing capabilities of the non-death-domain-containing receptor TNF-R2(Heller et al., 1992;Tartaglia et al., 1993), the use of agonistic TNF-R2-specific antibodies clearly showed that exclusive triggering of this receptor is sufficient in some cells to induce cell death(Bigda et al., 1994;Grell et al., 1993;Medvedev et al., 1994). However, stimulation of TNF-R2 does not directly engage the apoptotic program,but relies on the induction of endogenous TNF, which subsequently activates TNF-R1 (Grell et al., 1999;Vercammen et al., 1995). In addition, activation of TNF-R2 can lead to a tremendous enhancement of TNF-R1-induced cell death independent of endogenous TNF by a TRAF2-dependent intracellular mechanism (Chan and Lenardo,2000; Declercq et al.,1998; Duckett and Thompson,1997; Vandenabeele et al.,1995; Weiss et al.,1997).

Here, we show that stimulation of TNF-R2 results in a strong recruitment of TRAF2 that is accompanied by a strong depletion of cytosolic TRAF2. Moreover,we give evidence that TNF-R1 and TNF-R2 can compete for TRAF2-dependent recruitment of the anti-apoptotic proteins cIAP1 and cIAP2, which suggests that TNF-R2-induced TRAF2-mediated depletion of cIAP1/2 underlies the apoptotic TNF receptor crosstalk. In accordance with this model we show that TNF-R1-induced activation of procaspase-8 is slower than TNF-R2-induced TRAF2 depletion.

The human rhabdomyosarcoma cell line Kym1 was a gift from M. Sekiguchi(University of Tokyo, Japan) and maintained in Click-RPMI medium with 10%heat-inactivated fetal calf serum (Biochrom, Berlin, Germany). The human colon carcinoma cell line Colo205 and HeLa cells were from the American Type Culture Collection (Rockville, MD) and maintained in RPMI 1640 medium supplemented with 5% heat-inactivated fetal calf serum (Biochrom). The Jurkat-TNF-R2 transfectant was generously supplied by F. Pimentel-Muinos and B. Seed(Massachusetts General Hospital, MA) and maintained in RPMI 1640 medium with 10% heat-inactivated fetal calf serum(Pimentel-Muinos and Seed,1999). HeLa transfectants stably expressing TNF-R2 were as described elsewhere [HeLa-TNF-R2 (Weiss et al., 1997)]. HeLa-CD40 cells were a kind gift of H. Engelmann(University of Munich, Germany) (Hess and Engelmann, 1996).

For generation of HeLa-CD30 cells, HeLa cells were stably transfected with a CD30 expression plasmid. After G418 selection more than 100 primary clones were pooled, expanded and enriched for cells with high expression of cell surface CD30 by three rounds of cell sorting using a FACStar^Plus(Becton Dickinson, San Jose, CA) and the CD30-specific mAb Ki-1. Expression plasmids encoding TRAF2-GFP, TRAF2-YFP, TNF-R1-YFP, cIAP1(NT)-GFP and cIAP2(NT)-GFP were generated by proofreading PCR-based amplification of the corresponding cDNA regions from HeLa cDNA and insertion of the obtained amplicons into the pEGFP-N1 and pEYFP-N1 vectors (Clontech, Heidelberg,Germany). Forward primers contained a BamHI (BglII for cIAP1 plasmids; NheI for TNF-R1 plasmid) site and reverse primers contained a SacII (SacI for TNF-R1 plasmids) site in their 5′overhangs to allow oriented insertion into the respective sites of the pEGFP-N1 and pEYFP-N1 vectors.

The deletion mutant of TRAF2, where the C-terminal TRAF domain is substituted by GFP (TRAF2-NT-GFP), was generated by proofreading PCR based amplification of a cDNA fragment encoding amino acids 1-272 of TRAF2 and insertion of the obtained amplicon into pEGFP-N1. Again primers were used with BamHI and SacII overhangs to allow in frame insertion into the BglII and SacII sites of pEGFP-N1. To generate the N-terminal deletion mutant of TRAF2 in which the TRAF domain of the molecule was C-terminally fused to GFP, a cDNA fragment comprising TRAF2 amino acids 186-501 was amplified by proofreading PCR with HeLa cDNA as template. The forward primers used contained a BamHI site and the reverse primers used contained a NotI site. The BamHI and NotI sites were used for subsequent cloning of the amplicon in a modified version of the recently described pcDNA3.1-GFPΔFADD expression construct(Wajant et al., 1998). In this modified version a BamHI/BglII hybrid site between GFP andΔFADD was restored to a complete BamHI site, and a BamHI site 5′ to the start codon of the fusion protein was destroyed. Thus, in this modified version the ΔFADD part could be substituted by the respective TRAF2 amplicon.

The expression construct CFP-TRADD-DD was obtained by introducing a respective cDNA amplicon into the BglII and SacII sites of pECFP-N1. The expression vector pECFP-Mem encoding ECFP targeted to cellular membranes by 20 amino acids of neuromodulin was from Clontech (Heidelberg,Germany). Rabbit polyclonal anti-TNF-R2 IgG was already described elsewhere(Grell et al., 1995). Flag-tagged recombinant soluble TRAIL, FasL and CD40L were generously supplied by P. Schneider and J. Tschopp (University of Lausanne, Switzerland). If not otherwise stated, all other reagents were from Sigma (Deisenhofen,Germany).

Cells were seeded in 96-well microtiter plates (20×10³HeLa cells per well; 15×10³ Kym1 cells per well) and cultivated over night. The next day HeLa cells were treated with 2.5 μg/ml cycloheximide and Kym1 cells remained untreated. After 1 hour, TNF or FasL were added and costimulation of TNF-R2, CD30 or CD40 was performed as follows:TNF-R2 was stimulated with an agonistic TNF-R2-specific rabbit IgG fraction (2μg/ml); CD40 was stimulated with Flag-tagged recombinant soluble CD40L (100 ng/ml) crosslinked with the Flag-specific mAb M2 (1 μg/ml; Sigma); and CD30 was stimulated with the agonistic CD30-specific mAb Ki-1 (3 μg/ml). FasL was used as Flag-tagged recombinant soluble molecule crosslinked with 1μg/ml M2 (Schneider et al.,1998). After an additional 18 hours, cells were washed with PBS followed by crystal violet staining (20% methanol, 0.5% crystal violet) for 15 minutes. The wells were washed with H₂O and air-dried. The dye was eluted with methanol for 15 minutes and optical density at 550 nm was measured with a R5000 ELISA plate reader (Dynatech, Guernsey, GB).

HeLa-TNFR2 cells were seeded in 96-well plates with a density of 1.5×10⁴ cells per well. The next day cells were transfected with 200 ng/well of the indicated expression plasmid, 35 ng/well of a luciferase reporter plasmid driven by three consensus NF-κB sites and 15 ng/ml of β-galactosidase expression vector driven by the SV40 promoter using the Superfect reagent (Qiagen, Hilden, Germany) according to the manufacturer's instructions. After 1 day, cells were stimulated as indicated and cell extracts were prepared by the addition of 50 μl of luciferase lysis solution (Galactolight-Kit, Tropix, Bedford, MA) and one freeze-thaw cycle. A portion of the extracts (25 μl) was mixed with 50 μl of luciferase substrate (Luciferase Assay System, Promega) and the luminescence was determined in the single photon mode using an Anthos microplate luminometer (Lucy 2). In parallel, 25 μl of each cell extract was incubated for 1 hour with a 1:100 dilution of Galacton substrate in reaction buffer and mixed with 100 μl of accelerator II solution to determine relativeβ-galactosidase activity (Galactolight-Kit) again with the Anthos microplate luminometer (Lucy 2). Luciferase activities were normalized based on the β-galactosidase activities.

For detection of IκB, caspase-8 and caspase-3 pellets of 2×10⁶ cells were lysed in lysis buffer containing 150 mM NaCl, 50 mM Tris-HCl, pH 8.0, 1% NP-40 and 0.1% SDS, and lysates were clarified by centrifugation (4°C, 10 minutes, 13,000 rpm). Protein concentrations were determined using a Bradford based protein assay (Bio-Rad)and equal amounts of protein were resolved by SDS-polyacrylamide gel electrophoresis and transferred on nitrocellulose membranes. After blocking with 5% nonfat dry milk in PBS-Tween-20 (0.05%) for 1 hour, membranes were probed with the indicated antibodies. The antibody-antigen complexes were finally detected using alkaline phosphatase labeled secondary antibodies (0.1μg/ml, Sigma-Aldrich, Deisenhofen, Germany) and BCIP and NBT as substrates. For analyzing total cellular TRAF2 content by western blot, cell pellets were directly solubilized in Laemmli sample buffer (10⁶ cells per 50μl buffer) by help of a sonicator. The equivalent of 0.8×10⁶ cells per group was applied on the gel.

For preparation of TRAF2 cytoplasmic extracts and the corresponding detergent-insoluble fraction cells were lysed in 150 mM NaCl, 50 mM Tris-HCl pH 7.7, 1 mM EDTA and 1% Triton X-100 at 4°C for 1 hour. Then, the cytoplasmic fraction was cleared by centrifugation (4°C, 30 minutes,10,000 g). The obtained detergent-insoluble pellet was washed twice and solubilized in Laemmli sample buffer. Protein concentrations were determined using a Bradford based protein assay (Bio-Rad) and equal amounts of protein (100 μg cytoplasmic extract or the corresponding amount of insoluble pellet) were resolved by SDS-PAGE and analyzed by western blotting. Anti-caspase-3 mAb clone 19 was from Transduction Laboratories (Heidelberg,Germany) and anti-IκB was from Santa Cruz (Heidelberg, Germany). The anti-caspase-8 mAb was a kind gift from Klaus Schulze-Osthoff (University of Tübingen, Germany).

Cells (100-300×10⁶) were treated with the reagents of interest for the indicated times and subsequently scraped with a rubber policeman into the medium. Cells were washed in an ice-cold solution containing 50 mM Tris, pH 7.4, 400 mM NaCl and 10% glycerol, and the cell pellet was resupended in 0.5-1-fold its volume of the same ice-cold solution. All the following procedures were performed on ice or at 4°C. Cells were lysed with Nonident-P40 to a final concentration of 0.1% and a protease inhibitor cocktail (Boehringer Mannheim, Germany) was added according to the recommendations of the supplier. The lysates were cleared by centrifugation(8000 g, 10 minutes, 4°C) and the S-100 supernatants were obtained by centrifugation at 100,000 g for 1 hour in a TL-100 rotor at 4°C (Beckman, Munich, Germany). 200 μl of the S-100 supernatants were then separated by size exclusion chromatography on a Superdex 200 HR 10/30 column (Pharmacia, Freiburg, Germany) in 50 mM Tris, 400 mM NaCl and 10% glycerol, pH 7.4 at 0.5 ml/minute. Samples were collected in fractions of 0.5 ml and analyzed by immunoblotting with a polyclonal rabbit TRAF2-specific IgG fraction (Santa Cruz). The column was previously calibrated with thyroglobulin (663 kDa), apoferritin (443 kDa), alcohol dehydrogenase(150 kDa), bovine serum albumin (66 kDa), carbonic anhydrase (29 kDa) and cytochrome c (12.4 kDa), all purchased from Sigma.

Transfected cells were fixed onto coverslips in 3% paraformaldehyde,permeablized with 0.1% NP-40 for 10 minutes and blocked for a further 10 minutes with PBS/1% BSA. Primary rabbit antisera specific for Lck and TNF-R2 were diluted 1:150 in TBST (150 mM NaCl, 50 mM Tris-HCl pH 8.0, 0.1% Tween-20)and applied for 1 hour. Cells were then washed in PBS/0.05% Tween-20 for 15 minutes, incubated for 45 minutes with rabbit-specific antisera (diluted 1:200) conjugated with AlexaFluor 546 (A-11060), washed and mounted onto microscope slides. Alternatively, cells expressing GFP fusion proteins were maintained in a conditioned chamber (37°C, 5% CO₂) for up to 2 hours on the microscope stage. Fluorescent specimens were analyzed with a Leica SP2 confocal microscope and photographed using the Leica TCS software.

It has been shown that stimulation of TNF-R2(Weiss et al., 1997;Declercq et al., 1998;Chan and Lenardo, 2000) or CD40(Hess and Engelmann, 1996) as well as transient overexpression of a constitutive active form of CD30(Arch et al., 2000;Duckett and Thompson, 1997)enhance TNF-R1-induced cell death. To clarify whether these receptors use related mechanisms for the enhancement of TNF-R1-mediated cell death, HeLa transfectants stably overexpressing these receptors were analyzed(Fig. 1A). As shown inFig. 1B, costimulation of TNF-R2 and CD40 and, to a lesser extent, costimulation of CD30 enhanced TNF-R1-dependent cell death, which was induced by use of soluble TNF. There was no significant induction of apoptosis by selective stimulation of TNF-R2,CD30 and CD40 under the conditions used, excluding the possibility that these apoptotic effects are significantly based on the induction of endogenous TNF that had been described for these receptors in other systems(Grell et al., 1999;Vercammen et al., 1995). In all cases, induction of apoptosis by FasL and TRAIL, which both use the same FADD- and caspase-8-dependent pathway as that used by TNF-R1, was not affected by costimulation of these receptors (Fig. 1C,D). For TNF-R2 this specificity has already been described elsewhere (Weiss et al.,1997). TNF-R2-mediated enhancement of TNF-R1-induced cell death was also observed in Jurkat cells stably transfected with TNF-R2(Fig. 2A), as well as in Colo205 (Fig. 2B) and Kym1 cells (Fig. 2C). Together,these data point towards a principal mechanism by which TNF-R2, CD40 and CD30 affect formation of the TNF-R1-related death inducing signaling complex.

It has been shown that the transient overexpression of TNF-R2 or CD30 degraded co-transfected TRAF2 (Duckett and Thompson, 1997). Owing to the importance of TRAF2 for TNF-R1-mediated NF-κB activation(Devin et al., 2000;Devin et al., 2001;Tada et al., 2001),receptor-induced degradation of TRAF2 may account for the enhancement of TNF-R1-induced cell death by triggering of CD30, CD40 and TNF-R2. We therefore analyzed endogenous TRAF2 levels upon selective stimulation of TNF-R2 with agonistic mAbs. For this purpose we prepared S-100 supernatants from untreated HeLa-TNF-R2 cells and cells that were previously stimulated with TNF-R2-specific antibodies for 6 hours. Upon separation of the SN-100 supernatants over a superdex 200 HR10/30 column the fractions obtained were analyzed by immunoblotting. In accordance with data from the literature(Shu et al., 1996), we found that, in untreated cells, most of the cytosolic TRAF2 eluated in fractions corresponding to 300-500 kDa (Fig. 3A), suggesting that the majority of TRAF2 is part of a multiprotein complex in HeLa-TNF-R2 cells. But more importantly, when identical amounts of protein lysates were compared, obtained from receptor-stimulated and untreated cells, we found a significant depletion of TRAF2-containing complexes in the TNF-R2-stimulated group(Fig. 3B). TRAF2 depletion was most pronounced in the high molecular mass fractions(Fig. 3B), whereas a minority of TRAF2, which was eluted around 60 kDa, was not affected (data not shown). It is tempting to speculate that the `monomeric' TRAF2 fraction represents a pool of TRAF2 that is unable to interact with TNF-R2. However, further studies will be necessary to clarify the importance and function of this monomeric TRAF2 fraction. No TNF-R2-dependent TRAF2 depletion was found in HeLa cells stably expressing a deletion mutant of TNF-R2 lacking the cytoplasmic domain of the molecule (Fig. 3C).

In the stable transfectants expressing CD30 (HeLa-CD30) and CD40(HeLa-CD40), respectively, we also found a depletion of high molecular mass TRAF2-containing complexes upon stimulation with agonistic antibodies(Fig. 3D,E). In these cases we were unable to detect `monomeric' TRAF2 even in nonstimulated cells (data not shown). Compared with HeLa-TNF-R2 cells, we observed reduced levels of TRAF2 in unstimulated cells of both HeLa-CD30 and HeLa-CD40. This correlates with the observation that, in particular, HeLa-CD30 and, to a lesser extent,HeLa-CD40, but not TNF-R2-expressing cells exerted some ligand-independent constitutive receptor signaling (data not shown), possibly caused by the overexpression of these molecules. As `monomeric' TRAF2 represents a minor fraction of total TRAF2, it is possible that it was below the detection threshold of our western blot analysis. The `ligand'-induced depletion of TRAF2-containing complexes started to become apparent 1 hour after receptor triggering, and was almost complete after 6 hours(Fig. 3F). Depletion of TRAF2 occurred in the absence as well as in the presence of cycloheximide [i.e. under conditions where TNF-treatment could induce apoptosis in this cell line(data not shown)].

Receptor-induced depletion of cytoplasmic TRAF2-containing complexes could be caused by translocation of TRAF2 to a detergent-insoluble cellular compartment or/and by proteolytic degradation. We therefore analyzed total TRAF2 levels in cells by direct solubilization of cell pellets in SDS-PAGE sample buffer and determined TRAF2 levels in cytoplasmic extracts and detergent-resistant membrane fractions. Activation of TNF-R2 resulted in transient accumulation of TRAF2 in the detergent-resistant membrane fraction after 1-3 hours (Fig. 3G,H). However, 6-18 hours after receptor stimulation, a decrease in the TRAF content of total cell lysates was observed (Fig. 3G,H) suggesting that, after the first few hours of TNF-R2 stimulation, depletion of `free' cytoplasmic TRAF2 pools was mainly caused by recruitment of the protein by TNF-R2 into a detergent-insoluble compartment,followed by TRAF2 degradation leading to sustained downregulation of this signaling molecule. Nevertheless, TNF-R2-dependent depletion of TRAF2 also seems to continue at late time points, because the total content of TRAF2 dropped down by more than 60% (Fig. 3H) after 18 hours, whereas TRAF2 levels in the DIC fraction were comparable with those before stimulation(Fig. 3G). This implies that the distribution of TRAF2 between the soluble and the DIC fraction is still significantly shifted towards the DIC fraction. As an alternative to TRAF2 degradation, reduced de novo synthesis of TRAF2 could also explain the reduction of total TRAF2 protein, found at later time points, but we observed no changes in the TRAF2 mRNA content at any time after TNF-R2 stimulation(data not shown). In agreement with these findings Arch et al. have found that TRAF2 is redistributed from the cytoplasm into detergent-insoluble aggregates after co-transfection with constitutively active forms of CD30, 4-1BB and OX40(Arch et al., 2000). In this report redistribution of TRAF2 correlated with an increased sensitivity towards TNF. Reduction in cytoplasmic TRAF2 levels after CD30 stimulation was also associated with increased TNF sensitivity in anaplastic large cell lymphoma cells (Mir et al.,2000). In accordance with the idea that the availability of cytoplasmic TRAF2 determines the TNF sensitivity, it has been shown that HeLa transfectants stably expressing increased amounts of TRAF2 have a reduced sensitivity against soluble TNF but can still be sensitized to this death ligand by TNF-R2 stimulation (Weiss et al., 1998).

It has recently been shown that TRAF proteins can be the target of caspases[TRAF1 (Irmler et al., 2000),TRAF3 (Lee et al., 2001)] or the proteasome [TRAF6 (Takayanagi et al.,2000), TRAF2 (Brown et al.,2001)]. However, the ligand-induced depletion of TRAF2 observed in our system was not affected by the pan caspase-inhibitor z-VAD-fmk nor the proteasome inhibitor MG-132, even at concentrations tenfold higher than necessary to completely inhibit induction of apoptosis or activation of NF-κB, respectively (data not shown).

We next analyzed the effects of TNF-R2 stimulation and TRAF2 depletion on TNF-R1-induced signaling in detail. To analyze TNF-R1-induced NF-κB activation in a situation where TRAF2 was already depleted, we selectively prestimulated TNF-R2 with agonistic antibodies overnight, followed by TNF-R1 activation using either the TNF-R1-specific agonistic mAb HTR1 or soluble TNF. Under these conditions TNF-R1-mediated activation of a NF-κB-driven reporter gene was found to be reduced by up to 90% over a wide range of TNF(HTR1) concentrations (Fig. 4A,C). Comparable results were obtained in the presence of low doses of CHX (2.5 μg/ml) [i.e. under conditions where apoptosis is induced by TNF-R1 but protein synthesis is only modestly inhibited (data not shown)].

The situation was different when both TNF receptors were stimulated at the same time. Under these conditions there was no sign of TNF-R2-dependent inhibition of TNF-R1-induced NF-κB activation(Fig. 4B,D). Again, the presence of CHX did not change the outcome of the experiment (data not shown). As stimulation of TNF-R1 induces transient activation of the IKK complex within 5-15 minutes, which in turn leads to sustained NF-κB binding activity in the nucleus, it is feasible that the slowly proceeding TNF-R2-mediated TRAF2 depletion (Fig. 3) does not interfere with TNF-R1-induced NF-κB activation when both receptors are triggered at the same time. Indeed, the kinetics of TNF-R2-induced TRAF2 depletion (Fig. 3F) are in good agreement with the time dependency of NF-κB-inhibitory effect of TNF-R2 prestimulation. While TRAF2 depletion requires 1-3 hours, half maximal inhibition of TNF-R1-induced NF-κB activation by prestimulation of TNF-R2 became evident after about 2 hours(Fig. 4E). In accordance with the anti-apoptotic function of the NF-κB pathway, prestimulation of TNF-R2 for 8 hours resulted in a significantly greater enhancement of TNF-R1-induced cell death than a simultaneous stimulation of both TNF receptors (Fig. 4F). Nevertheless, as already shown in Figs1 and2, there is a clear enhancement of TNF-R1-induced cell death under costimulatory conditions where TNF-R1-induced NF-κB activation was not inhibited. Obviously, the apoptotic TNF-R1/TNF-R2 crosstalk is not only related to inhibition of the NF-κB pathway, but also based on an additional TNF-R2-dependent,NF-κB-independent mechanism, capable of enhancing TNF-R1-induced cell death at the post-transcriptional level.

In transient co-transfection studies it has been found that TRAF2 can be part of a complex containing cIAP1, cIAP2 and TRAF1, which inhibits TNF-R1-mediated apoptosis (Wang et al.,1998). TNF-R2-induced TRAF2 depletion/degradation could lead to inhibition of the formation of this anti-apoptotic complex and consequently to an enhancement of TNF-R1-induced activation of caspase-8. Under prestimulation conditions, where cytosolic TRAF2 is significantly depleted, the formation of the anti-apoptotic complex might be affected in two ways: (1) by reduced NF-κB-dependent induction of TRAF1, cIAP1 and cIAP2; and (2) by the lack of TRAF2 itself. In contrast, under costimulatory conditions only the lack of available TRAF2 may limit TRAF1/TRAF2/cIAP1/cIAP2 complex formation. However,when this is the case, the kinetics of TNF-R1-induced caspase-8 activation have to be slow or delayed to give the TNF-R2-dependent TRAF2 depletion time to take place and to interfere with the formation of, or to disrupt, the TRAF1/2/cIAP1/2 complex.

Slow or delayed caspase-8 activation by `activated' TNF-R1 is not self-evident. In the case of Fas-mediated apoptosis, a quantitative activation of procaspase-8 within a few minutes has been described(Medema et al., 1997). In HeLa cells treated with CHX, caspase-8 processing is detectable by cleavage of its full-length 55- and 53 kDa isoforms leading to disappearance of the respective proteins and concomitant appearance of the active p18 subunit of caspase-8.

As shown in Fig. 5A, almost all full-length caspase-8 was processed 1 hour after apoptosis-induction in HeLa-Fas cells by treatment with Fas-specific agonistic mAbs in the presence of CHX. In contrast, caspase-8 activation in TNF/CHX-treated cells was detectable only after 3 hours and complete activation was not achieved at all(Fig. 5A). Similar results were obtained by directly measuring caspase activity(Fig. 5B). The differential kinetics and extent of caspase-8 activation by stimulation of TNF-R1 and Fas were also observed in Jurkat, Kym1 (TNF-R1) and SKW (Fas) cells, where induction of apoptosis does not require the presence of CHX (data not shown). Similar results were obtained when the activity of the excutioner caspase-3 was monitored (Fig. 5A,B) or when crosslinked soluble FasL was used instead of Fas-specific agonistic antibodies (data not shown). The differences in kinetics and extent of caspase activation by TNF and anti-Apo1/FasL correlated with the variations in time,when the broad range caspase inhibitor z-VAD-fmk had to be added to prevent apoptosis. After 1 hour of Fas/Apo1 stimulation, z-VAD-fmk treatment was able to rescue only 60-70% of the cells and after 3 hours there was no protective effect at all (Fig. 5C). In contrast, upon TNF-R1 triggering a significant proportion of the challenged cells were protected by z-VAD-fmk even after 8 hours(Fig. 5C). Hence, under costimulatory conditions, the slow and ineffective activation of caspase-8 by TNF via TNF-R1 would allow TNF-R2-induced TRAF2 degradation to interfere with the action of the anti-apoptotic cIAP/TRAF complex. In accordance with this concept, we found that costimulation and, to an even greater extent,prestimulation of TNF-R2 enhanced TNF-R1-induced caspase-8 activation(Fig. 5A,B).

To investigate the role of TRAF2 for the recruitment of antiapoptotic factors into the TNF-R1 signaling complex we used confocal fluorescence microscopy and GFP-tagged forms of components of this complex. As shown inFig. 6A, cotransfection of TNF-R1-YFP and CFP-TRADD-DD led to a colocalization of both molecules. Interestingly, overexpression of CFP-TRADD-DD(Fig. 6A) or full-length myc-tagged TRADD (data not shown) resulted in the formation of TRADD filaments. As transiently overexpressed CFP-TRADD-DD and full-length myc-tagged TRADD showed effects similar to those of overexpressed TNF-R1 or ligand-stimulated TNF-R1, such as activation of the NF-κB(Fig. 6B), JNK and apoptosis pathways (data not shown) (Hsu et al.,1996), we postulate that TRADD filaments have some properties related to receptor-recruited endogenous TRADD. Importantly, we found that GFP-tagged IKK1, a part of the NF-κB-inducing IKK complex, was also partially recruited into the TRADD filaments in the presence of cotransfected TRAF2 (Fig. 6D). This is in accordance with a recent study using TRAF2^-/- fibroblasts, which showed that TRAF2 is necessary to recruit the IKK complex into the TNF-R1 signaling complex (Devin et al.,2001). Again this emphasizes our assumption that TRADD filaments are functionally equivalent to TNF/TNF-R1/TRADD signaling complexes.

We therefore used the artificial filament formation of TRADD to investigate the interaction of TRADD with TRAF2, cIAP1 and cIAP2. We first studied the interaction of untagged TRADD and GFP fusion protein of TRAF2. N-terminally as well as C-terminally GFP-linked TRAF2 behaved similarly to untagged TRAF2 in respect to homo- and heterooligomerization, NF-κB activation and binding of a TNF-R2 GST fusion protein (data not shown). In unstimulated cells, TRAF2 N-terminally linked to GFP (TRAF2-GFP) was found in the cytoplasm(Fig. 6C). In addition,TRAF2-GFP accumulates in a few, large round patches(Fig. 6C). Most likely these TRAF2-GFP patches are artefacts caused by overexpression and may represent the proportion of transiently expressed TRAF2 molecules that is responsible for ligand-independent activation of NF-κB and JNK. We found that TRAF2-GFP was recruited into TRADD filaments under critical involvement of its TRAF domain (Fig. 6C). Important for the idea that TRAF2 depletion by non-death receptors interferes with the formation/action of TNF-R1/TRADD-recruitable anti-apoptotic cIAP/TRAF-complexes, we found that TRAF2 dramatically enhanced the recruitment of cIAP1-GFP and cIAP2-GFP into TRADD filaments. While in cells coexpressing TRADD and cIAP1-GFP or cIAP2-GFP, only a small proportion of the GFP-tagged IAP proteins were associated with the TRADD filaments, coexpression of TRAF2 leads to a quantitative recruitment of cIAP1-GFP and cIAP2-GFP into TRADD complexes (Fig. 6E,F). Obviously, TRAF2 has an essential role in the recruitment of cIAP1 and cIAP2 into the TNF-R1/TRADD complex, data that are in good accordance with the postulated importance of the TRAF1/TRAF2/cIAP1/cIAP2 complex discussed above and with previous data showing ligand-induced recruitment of cIAP1 into the TNF-R1 signaling complex (Shu et al.,1996).

Next we studied directly alterations of the intracellular localization of GFP-linked TRAF2 induced by stimulation of TNF-R2. Real-time monitoring of TRAF2-GFP by confocal laser microscopy showed that, in HeLa-TNF-R2 cells,TRAF2-GFP becomes compartmentalized in small aggregates in response to selective activation of TNF-R2 (Fig. 7A), whereas Ki-1 or a control IgG preparation had no effect on HeLa-TNF-R2 cells (Fig. 7B,C). Similar results were obtained upon stimulation of CD30 or CD40. The TNF-R2-induced TRAF2-GFP containing aggregates represent complexes of TNF-R2 and TRAF2-GFP, as is evident from colocalization with the plasma membrane marker CFP-Mem (Fig. 7D) and TNF-R2-specific antibodies (Fig. 7E). Of particular interest, soluble TNF, which is unable to properly activate TNF-R2 in HeLa-TNFR2 cells, failed to induce receptor/TRAF2-GFP complexes (Fig. 9B). Importantly, cIAP1-GFP and cIAP2-GFP (data not shown) as well as GFP fusion proteins of the IAP1/2 binding domain for TRAF2(Fig. 8) were recruited into the TNF-R2 signaling complex only when TRAF2 was coexpressed. This is in accordance with data from the literature showing ligand-independent interaction of TNF-R2 and cIAP proteins in the presence of TRAF2 in transient overexpression experiments (Rothe et al.,1995). Analysis of cytoplasmic cIAP1-GFP and cIAP2-GFP revealed a significant depletion of these molecules 30 minutes after TNF-R2 stimulation(Table 1). In six independent experiments, reduction of cytoplasmic cIAP-GFP proteins ranged between 45 and 81% (average 52% for cIAP1-GFP and 56% for cIAP2-GFP. Depletion of IAP1/2-GFP in the presence of TRAF2 was typically more efficient than depletion of TRAF2-GFP (Table 1). Thus, the possibility arises that TRAF2/cIAP complexes interact better with TNF-R2 than with IAP-free TRAF2 complexes. However, it cannot be ruled out that this is caused by higher expression of TRAF2-YFP compared with the non-tagged TRAF2 mediating TNF-R2-recruitment of TRAF2/cIAP-GFP complexes in the IAP depletion experiments. Further experiments will be necessary to clarify this.

Based on the presented results we suggest the following model(Fig. 9) for the enhancement of TNF-R1-induced cell death by stimulation of non-death-domain-containing members of the TNF receptor superfamily that interact with TRAF2.

Exclusive stimulation of TNF-R1 (Fig. 9A) with soluble TNF immediately activates the NF-κB pathway leading to the synthesis of NF-κB-dependent genes. However, the formation of TNF-R1 complexes that can signal apoptosis occurs with a delay of several hours. This is evident from the time course of TNF-induced caspase-8 activation (Fig. 5). TNF-induced NF-κB activation results in the upregulation of anti-apoptotic factors such as TRAF1(Schwenzer et al., 1999;Wang et al., 1998), cIAP1(Wang et al., 1998), cIAP2(Chu et al., 1997;Wang et al., 1998), cFLIP(Kreuz et al., 2001), and Bfl-1/A1 (Lee et al., 1999). An anti-apoptotic function has been published for all these proteins and, with the exception of Bfl1/A1, all these molecules can interact with TRAF2. With respect to the apoptotic TNF-R1/TNF-R2 crosstalk, Bfl1/A1 can most likely be neglected as it acts independently of TRAF2, is barely detectable in HeLa cells and is not induced by TNF in this cell line (data not shown). Although cFLIP can bind to TRAF2 it is certainly not directly involved in the apoptotic TNF-R1/TNF-R2 crosstalk for several reasons: (1) In Jurkat-TNF-R2 cells, in which the apoptotic crosstalk occurs in the absence of CHX, cFLIP is not detectable and is not induced by TNF. (2) In many cell lines, including HeLa cells, which have been used in this study, at least low concentrations of CHX are necessary for the induction of death receptor-induced apoptosis. It is commonly assumed that this reflects the necessity to downregulate an anti-apoptotic factor with rapid turnover below a critical threshold concentration. A prime candidate for this anti-apoptotic factor is cFLIP, as we and others have recently found that cFLIP has a rapid turnover and is therefore its expression is highly sensitive to metabolic inhibitors such actinomycin D or CHX (Griffith et al.,1998; Kreuz et al.,2001; Leverkus et al.,2000; Wajant et al.,2000). It is noteworthy that the requirement for this compound for induction of apoptosis remains unchanged upon prestimulation of TNF-R2,indicating that this CHX-sensitive factor is distinct from the NF-κB-dependent anti-apoptotic factors mediating the apoptotic TNF-R1/TNF-R2 crosstalk. Moreover, the apoptotic TNF-R1/TNF-R2 crosstalk is FLIP-independent, which is in agreement with the facts that TNF- but not Fas-and TRAIL-induced apoptosis are enhanced by prestimulation of TNF-R2, and that apoptosis induction in Hela cells by all these reagents requires CHX. Therefore, according to our model, we suggest that the anti-apoptotic factors cIAP1, cIAP2, and TRAF1 regulate the apoptotic TNF-R1/TNF-R2 crosstalk. All these factors bind to TRAF2 and become recruited into the TNF-R1 signaling complex, thereby preventing efficient activation of caspase-8. The outcome is a modest apoptotic response. Hence, the balance between protective mechanisms and activation of caspase-8, the trigger of the apoptotic program, is shifted towards apoptosis only in a small fraction of the cells. cIAP1 and cIAP2 bind and inhibit caspase-3 and caspase-7 but not caspase-8(Roy et al., 1997). Thus, it is tempting to speculate that in the TNF-R1 signaling complex TRAF1/2-recruited IAPs and TRADD/FADD-recruited caspase-8 come into close proximity, which allows IAP proteins to inhibit caspase-8 activation, possibly without a direct interaction between the two. In accordance with the scenario described above, Wang et al. have found in NF-κB-activation-deficient p65^-/- fibroblasts that the individual transient overexpression of TRAF1, TRAF2, cIAP1 and cIAP2 has no significant inhibitory effect on TNF-induced apoptosis, whereas the coexpression of these molecules protects the cells from TNF-induced cell death(Wang et al., 1998).

The consequences of TNF-R1 stimulation, as described above, are quite different when TNF-R1 becomes triggered several hours after activation of TNF-R2 (Fig. 9B): prestimulated TNF-R2 has then depleted TRAF2. This has two consequences for TNF-R1 signaling. First, owing to the important role of TRAF2 in TNF-R1-induced NF-κB activation (Devin et al.,2000; Devin et al.,2001; Tada et al.,2001), this pathway is inhibited(Fig. 4) and therefore the production of anti-apoptotic factors is reduced. Second, low levels of pre-existing or still reduced levels of induced anti-apoptotic proteins cannot work properly as they need the support of TRAF2 [e.g. for targeting to the TNF-R1 signaling complex (Fig. 6)]. Reduced NF-κB activation and impaired formation of TRAF1/TRAF2/cIAP1/cIAP2 complexes fails to inhibit caspase-8 activation and the apoptotic pathway becomes activated in all cells.

The situation changes again when TNF-R1 is stimulated together with TNF-R2(Fig. 9C). TNF-R1-induced NF-κB activation per se is not inhibited under these conditions(Fig. 4), as this process is clearly faster than depletion/degradation of TRAF2. Hence, TNF-R1 triggering leads to normal production of anti-apoptotic factors. However, the biosynthesis of these factors needs time, and during this time TRAF2 is depleted/degraded by TNF-R2. Thus, although anti-apoptotic factors are produced they cannot act properly as they need TRAF2 for their action at the post-transcriptional level (see above). The outcome is an enhanced but,compared with the prestimulation scenario, reduced sensitization of the cells for the apoptotic effects of TNF-R1.

All the experiments described in this study and all experiments related to TNF receptor cooperation and TRAF2 depletion described in the literature were performed in cancer cell lines. Thus, the question of the physiological relevance of TRAF2 depletion and TNF receptor crosstalk was raised. In this regard, the available literature points to at least two physiological processes that could be related to this TNF receptor cooperation. First,induction of activation-induced cell death in CD8⁺ T cells. In accordance with an involvement of TRAF2 depletion and TNF receptor cooperation, it has been shown in TNF-R1(Speiser et al., 1996) and TNF-R2 (Zheng et al., 1995)knockout mice that both receptors can have a major impact on this TNF-mediated apoptotic process. Moreover, Sarin et al. have shown that TNF-R1 and TNF-R2 synergistically induce cell death in T cell blasts in vitro(Sarin et al., 1995). Second,TNF-R2-induced TRAF2 depletion and inhibition of TNF-R1-induced NF-κB activation under non-apoptotic conditions(Fig. 4) could explain the finding that in a mylelin oligodendrocyte glycoprotein (MOG) model of experimental autoimmune encephalomyelitis (EAE) TNF- or TNF-R1-deficiency led to delay in disease onset, whereas TNF-R2-deficiency results in a more severe phenotype (Eugster et al.,1999).

This work was supported by Deutsche Forschungsgemeinschaft Grant Wa 1025/3-1 and Sonderforschungsbereich 495 project A5.