Analysis of CD95 Threshold Signaling

Recently we generated a mathematical model (Bentele, M., Lavrik, I., Ulrich, M., Stosser, S., Heermann, D. W., Kalthoff, H., Krammer, P. H., and Eils, R. (2004) J. Cell Biol. 166, 839-851) of signaling in CD95(Fas/APO-1)-mediated apoptosis. Mathematical modeling in combination with experimental data provided new insights into CD95-mediated apoptosis and allowed us to establish a threshold mechanism of life and death. Here, we further assessed the predictability of the model experimentally by a detailed analysis of the threshold behavior of CD95 signaling. Using the model predictions for the mechanism of the threshold behavior we found that the CD95 DISC (death-inducing signaling complex) is formed at the cell membrane upon stimulation with low concentrations of agonistic anti-APO-1 monoclonal antibodies; however, activation of procaspase-8 at the DISC is blocked due to high cellular FLICE-inhibitory protein recruitment into the DISC. Given that death signaling does not occur upon CD95 stimulation at low (threshold) anti-APO-1 concentrations, we also analyzed survival signaling, focusing on mitogen-activated protein kinase activation. Interestingly, we found that mitogen-activated protein kinase activation takes place under threshold conditions. These findings show that triggering of CD95 can signal both life or death, depending on the strength of the stimulus.

ized, probably trimerized, CD95 receptors, the adaptor molecule FADD, two isoforms of procaspase-8 (procaspase-8/a and procaspase-8/b), procaspase-10, and c-FLIP L/S/R (6,10). The interactions between the molecules at the DISC are based on homotypic contacts. The death domain (DD) of the receptor interacts with the DD of FADD, while the death effector domain (DED) of FADD interacts with the N-terminal tandem DEDs of procaspases-8, -10, and c-FLIP L/S/R . The binding of procaspase-8 to the DISC results in a two-step processing of the zymogen (11). The first cleavage step generates the two subunits p43/p41 and p12. In a second cleavage step the active enzyme subunits p18, p10, and the prodomains p26/p24 are produced. As a result, the active caspase-8 heterotetramer p10 2 -p18 2 is released into the cytosol to propagate the apoptotic signal (12).
Two CD95 signaling pathways were established (13). Type I cells are characterized by high levels of CD95 DISC formation and increased amounts of active caspase-8. Activated caspase-8 directly leads to the activation of downstream effector caspases-3 and -7. Type II cells are characterized by lower levels of CD95 DISC formation and, thus, lower levels of active caspase-8 (13). In this case, signaling requires an additional amplification loop that involves the cleavage by caspase-8 of the Bcl-2 family protein Bid to generate truncated (t) Bid and subsequent tBid-mediated release of cytochrome c from mitochondria. The release of cytochrome c from mitochondria results in apoptosome formation, followed by activation of procaspase-9, which in turn cleaves downstream, effector caspases (14).
Recently there have been a few reports demonstrating that CD95 is not only a potent apoptosis inducer but is also capable of activating multiple survival pathways (15)(16)(17)(18). CD95 stimulation was shown to play a central role in neurite outgrowth and neuronal regeneration following injury (17). This effect was shown to involve the activation of the mitogen-activated protein (MAP) kinases ERK1/2 (17). In tumor cells resistant to CD95-induced apoptosis (e.g. MCF7-Fas-Bcl-x L ) the triggering of CD95 was reported to result in activation of survival pathways involving NF-B, ERK1/2, p38, c-Jun N-terminal kinase (JNK), and AKT (16). However, the exact molecular mechanism of survival signaling via CD95 remains unclear.
To gain a better understanding of CD95-induced apoptosis and to elucidate the decision-making steps of CD95 signaling we developed a mathematical model of this signaling pathway (1). To estimate the parameters of the model we performed quantitative measurements of CD95-induced apoptosis in human B lymphoblastoid SKW6.4 cells. SKW6.4 cells are Type I cells that are highly sensitive to CD95-mediated apoptosis. The model made a number of important predictions on the mechanisms of life/death decisions in CD95-induced apoptosis. According to these predictions CD95 signaling is characterized by a threshold behavior. The model predicted that the threshold is defined at the DISC and depends on c-FLIP concentrations at the DISC (10,19).
In this study using the model predictions we address the question how initial events in CD95 signaling influence life/ death decisions at the CD95. We show that the initial concentration of CD95L and the concentration of the inhibitor of caspase-8 c-FLIP at the DISC are the most important factors contributing to the cellular decisions for CD95-induced life and death.
Flow Cytometry Analysis-The percentage of viable cells was determined by forward scatter/SSC using a FACSscan cytometer (BD Biosciences). A minimum of 10,000 cells/sample was analyzed. Specific cell death was calculated as follows: (percentage of experimental cell death Ϫ percentage of spontaneous cell death)/(100 Ϫ percentage of spontaneous cell death) ϫ 100.
Western blots were quantified using the LumiImager TMF1 system and Lumi Analyst software Version 3.0 (Roche Applied Science). The measurements were always performed in the region of a linear relation between the amount of antigen and the signal strength. The light signal was measured in Boehringer light units and in relative amounts. The Boehringer light unit value calculated is the absolute integration value of the band that was evaluated. The S.D. was calculated from up to three independent experiments.
Caspase Activity Assays-Cytosolic lysates were incubated with 50 M site-specific tetrapeptide substrates (zIETD-afc for caspase-8, zDEVD-afc for caspase-3) in caspase assay buffer B (50 mM HEPES, 100 mM NaCl, 10 mM dithiothreitol, 0.1% (w/v) CHAPS, 10% (w/v) sucrose, pH 7.4) in a final volume of 200 l. The release of the fluorogenic group AFC was determined after 1 h of incubation at 37°C by a microplate fluorescence reader Wallach 1420 (PerkinElmer Life Sciences) at the excitation wavelength of 405 nm and emission wavelength of 535 nm.
DISC Analysis by Immunoprecipitation and Western Blotting-1 ϫ 10 8 cells in a concentration of 10 6 cells/ml were treated with indicated concentration of anti-APO-1 (IgG3) for 10 min at 37°C, washed twice in 1ϫ phosphate-buffered saline (PBS), and subsequently lysed in buffer A (20 mM Tris/HCl, pH 7.5, 150 mM NaCl, 2 mM EDTA, 1 mM phenylmethylsulfonyl fluoride (Sigma), protease inhibitor mixture (Roche Applied Science), 1% Triton X-100 (Serva), and 10% glycerol) (stimulated condition) or lysed without treatment (unstimulated condition). The CD95 DISC was immunoprecipitated overnight with 2 g of anti-APO-1 and Protein A-Sepharose. Beads were washed five times with 20 volumes of lysis buffer. The immunoprecipitates were analyzed on 12% polyaerylamide gel. Subsequently, the gels were transferred to Hybond nitrocellulose membrane (Amersham Biosciences), blocked with 5% nonfat dry milk in PBS/Tween (PBS plus 0.05% Tween 20) for 1 h, washed with PBS/Tween, and incubated with the primary antibodies in PBS/Tween at 4°C overnight. Blots were developed with a chemoluminescence method following the manufacturer's protocol (PerkinElmer).
Caspase Inhibition Experiments-Z-VAD-fmk (pan-specific caspase inhibitor) was added to the cells in the final concentration of 50 M 30 min before the addition of anti-APO-1.
Proliferation Assays-For proliferation assays 1 ϫ 10 5 cells were seeded into a 96-well titer plate and stimulated with either 1 ng/ml of anti-APO-1 or left untreated for up to 4 days. Proliferation was measured with a scintillation counter after tritiated thymidine ([ 3 H]TdR) incorporation during the final 15-18 h of the culture.

CD95 Stimulation with Low Concentrations of Anti-APO-1 Antibodies Results neither in Cell Death nor in Caspase
Activation-The analysis of CD95 signaling in human B lymphoblastoid SKW6.4 cells using mathematical modeling in combination with biochemical approaches led to the finding of a threshold behavior of CD95-induced apoptosis (1). We established the threshold concentration of anti-APO-1 antibodies that did not result in cell death in SKW6.4 cells to be 1 ng/ml (1). We performed a more detailed analysis of CD95-mediated cell death in SKW6.4 cells than in our previous studies (1) (Fig. 1). We varied the concentration of anti-APO-1 from 0.1 ng/ml to 1 g/ml and measured cell death at different time points from 2 h to 4 days (Fig. 1A). Notably, we used the same concentration of MAY 4, 2007 • VOLUME 282 • NUMBER 18

JOURNAL OF BIOLOGICAL CHEMISTRY 13665
SKW6.4 cells (10 6 cells/ml) as in our previous studies (1). Upon stimulation with 1 ng/ml of anti-APO-1 antibodies (threshold concentration) we did not observe any cell death (Fig. 1A) nor did we see caspase activation using the fluorescent substrates of caspases zIETD-afc and zDEVD-afc (Fig. 1B).
To assure that the threshold concentration of the anti-CD95 stimulus does not lead to an initial low level of caspase activation, which would disappear at a later time point, we analyzed caspase cleavage by Western blot in the first hours after stimu-lation (Fig. 1C). SKW6.4 cells (10 6 cells/ml) were stimulated with 1 ng/ml (threshold concentration) or with 100 ng/ml of anti-APO-1 antibodies. No processing products of procaspase-8, procaspase-3, or poly(ADP-ribose) polymerase cleavage were observed upon stimulation with the threshold concentration of 1 ng/ml of anti-APO-1 (Fig. 1C). In contrast, upon stimulation with 100 ng/ml we observed cleavage of procaspase-8, procaspase-3, and poly(ADP-ribose) polymerase, indicating triggering of the CD95-induced apoptosis. Thus, we reconfirmed the threshold concentration of anti-APO-1 (1 ng/ml) for SKW6.4 cells and further characterized the threshold behavior of these cells showing that caspases are not activated within the first hours after stimulation.
The CD95 DISC Is Formed at Threshold Concentrations in Lower Amounts and c-FLIP Inhibits Caspase-8 Activation at the DISC upon Threshold Stimulation-Next, we tested the threshold mechanism at the DISC. Our model has made the following predictions for the mechanism of threshold behavior of CD95 signaling: lower amounts of CD95 DISCs, high c-FLIP recruitment to the DISC due to the highest affinity of c-FLIP to the DISC compared with other DED proteins (21), and, consequently, inhibition of caspase-8 activation (1) (Fig. 2A).
Our model suggested that upon threshold stimulation the CD95 DISC is formed, albeit in low amounts (1). So far we had not analyzed CD95 DISC formation and composition under threshold concentrations of anti-APO-1 antibodies. To address this issue we performed CD95 immunoprecipitations upon stimulations with 1, 10, and 100 ng/ml of anti-APO-1 antibodies (Fig. 2B). We used the same concentration of SKW6.4 cells (10 6 cells/ml) to strictly follow the threshold conditions for these cells established in Fig. 1 and Ref. 1. After stimulation the CD95 DISC was immunoprecipitated by using only Protein A-Sepharose without the addition of anti-APO-1 antibodies (Fig. 2B). Importantly, we observed that the CD95 DISC is indeed formed upon stimulation with 1 ng/ml of anti-APO-1 antibodies. FADD, procaspase-8a/b, and c-FLIP L/R molecules were recruited to the DISC, albeit in lower amounts, as compared with the DISC formed upon stimulation with 100 ng/ml. As anticipated, we observed a prominent decrease of the receptors engaged in DISC formation upon stimulation from 100 to 1 ng/ml of anti-APO-1. Thus, we confirmed the first statement of the  (1). Upon lower amount of stimulus (anti-APO-1 antibodies) the amount of active DISC binding sites is very low (red line). As the DISC binding sites we considered the free FADD molecules at the DISC available for the binding of DED-containing molecules (procaspase-8, c-FLIP). Active DISC binding sites are quickly occupied by c-FLIP molecules and, correspondingly, are blocked for procaspase-8 binding and activation (green line). B, SKW6.4 cells were stimulated with 1, 10, or 100 ng/ml of anti-APO-1 antibodies for 10 min. CD95 DISC was immunoprecipitated using only Protein A-Sepharose. Immunoprecipitated DISC components CD95, FADD, c-FLIP L/R , procaspase-8 (p55/p53), and its cleavage products p43/p41 and p18 were analyzed by Western blot. C, CD95 was immunoprecipitated from SKW6.4 cells either stimulated with 10 ng/ml of anti-APO-1 or left untreated. For unstimulated cells the immunoprecipitation was performed using 10 or 1 ng of anti-APO-1. Immunoprecipitated c-FLIP and procaspase-8 were analyzed by Western blot. D, SKW6.4 cells were stimulated with 100 ng/ml of anti-APO-1 with pretreatment with 50 M Z-VAD-fmk. CD95 DISC was immunoprecipitated and loaded on one gel along with 1/20 amount of total cellular lysates, from which immunoprecipitation was performed. Immunoprecipitated DISC components CD95, FADD, c-FLIP L/R , procaspase-8 (p55/p53), and its cleavage products p43/p41 and p18 were analyzed by Western blot. Relative recruitment of c-FLIP and procaspase-8 to the DISC from the lysates was calculated using quantitative Western blot. We considered as the amount of procaspase-8 in one lane the sum of the signal corresponding to the bands of procaspase-8 (p55/p53) and its cleavage product p43/p41. As the amount of c-FLIP in one lane we summarized the signals resulting from c-FLIP L (p55), its cleavage product (p43), and c-FLIP R . model that the CD95 DISC is formed upon threshold concentrations of anti-APO-1, albeit in lower amounts (1).
Interestingly, in Fig. 2B we also observed that the amount of c-FLIP recruited to the DISC was higher for stimulation with 1 ng/ml than with 100 ng/ml. This result confirms the next prediction of the threshold mechanism.
Importantly, the processing of procaspase-8 at the DISC upon threshold concentrations also confirmed the model predictions (Fig. 2B). We observed that upon stimulation with 1 ng/ml procaspase-8 is processed only to p43/p41, indicating that processing of procaspase-8 is stopped after the first cleavage step. In contrast, for the stimulation with 100 ng/ml we also observed the generation of the p18 subunit, indicating that for the higher concentrations of anti-APO-1 the second cleavage step takes place and active caspase-8 is released into the cytosol to start the apoptotic signaling cascade.
To rule out that under threshold stimulation of 1 ng/ml we immunoprecipitate low amounts of DISCs, which are spontaneously formed in the cells without stimulation, we performed immunoprecipitations from unstimulated cells with 1 and 10 ng/ml of anti-APO-1 (Fig. 2C). Procaspase-8 and c-FLIP were not associated to CD95 under these conditions.
As the affinities of procaspase-8 and c-FLIP to the DISC are critical parameters for the decision making at the DISC, we compared the amounts of procaspase-8 and c-FLIP that were recruited to the DISC from corresponding lysates by quantitative Western blotting (Fig. 2D). For each molecule, e.g. procaspase-8 or c-FLIP, we measured their amounts at the DISC versus their amounts in the lysates. To prevent proteolytical processing of procaspase-8 to p26/p24, which is not detectable by anticaspase-8 antibody directed against the C terminus of procaspase-8, we added pan-caspase inhibitor Z-VAD-fmk. Under these conditions we observed that the amount of c-FLIP at the DISC was eight times higher than the amount of c-FLIP in the lysates. The amount of procaspase-8 at the DISC was five times higher that the amount of procaspase-8 in the lysates. Thus, the affinity of c-FLIP to the DISC was roughly twice as high compared with that of procaspase-8. Thus, we showed that c-FLIP has a higher affinity to the DISC than procaspase-8.
We could not compare the ratio of procaspase-8 to c-FLIP at the DISC due to the fact that we were using different antibodies. The question of stoichiometry of the DISC is a matter for future studies. Thus, we have confirmed our model predictions and show that at threshold stimulation c-FLIP is up-regulated at the DISC due to the high affinity of c-FLIP to the DISC, leading to the inhibition of caspase-8 activation and further downstream apoptotic events.
Triggering of CD95 with Low Amounts of a CD95 Stimulus Does Not Result in Apoptosis but in MAP Kinase Activation-Next we analyzed whether stimulation of CD95 with threshold amounts of anti-APO-1 antibodies triggers reported survival pathways (15)(16)(17)(18). We observed that stimu- lation of SKW6.4 cells with 100 ng/ml of anti-APO-1 antibodies resulted in activation of ERK and p38 (Fig. 3A). Importantly, this parallels procaspase-8 processing (Fig. 3A), resulting in cell death (Fig. 1). The stimulation with isotype control antibodies (FII23C) did not lead to MAPK activation (Fig. 3B). Pretreatment of SKW6.4 cells with the pan-caspase inhibitor Z-VAD-fmk did not inhibit MAP kinase activation (Fig. 3C), indicating that CD95-mediated MAP kinase activation is caspase-independent. Interestingly, SKW6.4 cells are highly sensitive to CD95-mediated apoptosis; however, triggering of CD95 also leads to survival signaling apparently in parallel with death signaling.
The decrease of concentrations of anti-APO-1 antibodies down to the threshold level (1 ng/ml) did not result in blockage of MAP kinase activation (Fig. 4A). However, we observed clear differences in the kinetics of ERK phosphorylation upon stimulation with different concentrations of anti-APO-1. Upon stimulation with high concentrations of anti-APO-1 pERK peaked earlier than for the stimulation with 1 ng/ml of anti-APO-1 antibodies. To address this issue in more detail we performed several new experiments of CD95mediated pERK activation (Fig. 4B). We observed an increase in ERK phosphorylation for all concentrations used; however, the phosphorylation peaked at different time points in different experiments. This probably reflects the different initial status of ERK phosphorylation of the cells used for our experiments.
To study whether threshold activation leads to engagement of other life pathways we looked at NF-B activation, using as a readout the phosphorylation of IB␣ (Fig. 4C). Indeed, we observed phosphorylation of IB␣ upon stimulation with 1 ng/ml of anti-APO-1, even though the kinetics of phosphorylation were different from the ones when cells were stimulated with higher concentrations of anti-APO-1.
To understand the physiological role of stimulations with low concentration of anti-APO-1 we performed proliferation assays in SKW6.4 cells upon stimulation with 1 ng/ml of anti-APO-1 (data not shown). However, we did not find a significant increase in proliferation rates in SKW6.4 cells. Because SKW6.4 are fast proliferating cancer cells it might be of advantage to establish a primary cell system to have a better readout of the increase in proliferation. This is a matter for future work. Thus, we observed that when cell death is blocked MAP kinase activation still takes place, indicating that triggering of CD95 with low concentrations of stimulus results only in survival signaling.

DISCUSSION
In this study using the model predictions we address the question how initial events in CD95 signaling influence life/ death decisions at the CD95. We showed that the initial concentration of CD95L and the concentration of the inhibitor of caspases c-FLIP are the most important factors contributing to the cellular decisions for CD95-induced life and death. We demonstrated that upon threshold stimulation 1) the CD95 DISC is formed, albeit in low amounts; 2) c-FLIP is quickly recruited to the DISC due to its high affinity to the DISC; and 3) c-FLIP inhibits caspase-8 activation by blocking the second step of procaspase-8 processing at the DISC, resulting only in the formation of p43/p41 subunits.
We performed our study using only SKW6.4 cells and anti-APO-1 antibodies as we quantitatively characterized CD95-induced apoptosis in SKW6.4 cells in our previous work and experimentally determined the threshold concentration of anti-APO-1 antibodies in SKW6.4 cells (1). This quantitative characterization allowed us to proceed directly to more detailed DISC analysis in the present study.
The central molecule defining the threshold behavior in CD95-induced apoptosis is c-FLIP (1). c-FLIP has a dual function in CD95-mediated apoptosis. c-FLIP S/R is a devoted inhibitor of CD95-induced apoptosis, whereas c-FLIP L is described to play both pro-and anti-apoptotic roles (10,19,21,22). When present in the cells in low concentrations, c-FLIP L accelerates procaspase-8 activation at the DISC (21). Upon high concen-trations in the cells (e.g. overexpression conditions), c-FLIP L was reported to inhibit CD95-induced apoptosis (19,21). The mechanism of c-FLIP L inhibition involves blocking the second step of procaspase-8 processing (19). We observed that at threshold concentrations of anti-APO-1 procaspase-8 processing at the DISC is also stopped after the first cleavage step with the formation of p43/p41 subunits (Fig. 2B). These findings indicate that inhibition of procaspase-8 processing at the DISC upon threshold stimulation is similar to the one seen when c-FLIP is up-regulated. Indeed, we observed that amount of c-FLIP at the DISC at threshold concentrations of anti-APO-1 is higher than the one for high concentrations of anti-APO-1 (Fig. 2B). This should lead to the inhibition of the second step of caspase-8 processing and, thereby, apoptosis (Fig. 5).
In addition to inhibition of the second cleavage step of procaspase-8 at the "threshold" DISC, we also observed inhibition of the first cleavage step as judged by reduced amounts of p43/ p41 (Fig. 2B). This might be explained by up-regulation of c-FLIP R at the threshold DISC blocking the first cleavage step. Upon stimulation CD95 can trigger both survival and the death pathways. The signal transduction for the death pathway was characterized in detail and involves DISC formation with subsequent activation of procaspase-8 at the DISC (left side). Activation of procaspase-8 involves two cleavage steps, p43/p41 and p10 generation, which is followed by the second step with formation of p18 and prodomain. Active caspase-8 heterotetramer p10 2 -p18 2 goes to the cytosol to propagate the apoptotic signal. When triggered with high concentrations of CD95 stimulus the death signaling is predominant. At threshold concentrations of CD95 stimuli (right side) procaspase-8 processing at the DISC is stopped after the first step (formation of p43/p41 and p10). Thus, apoptosis is blocked and "life" signaling is predominant.
To get more insights into the mechanism of inhibition of procaspase-8 at the DISC upon threshold stimulation we compared the affinities of c-FLIP to procaspase-8 at the DISC (Fig.  2D). We found that affinity of c-FLIP is indeed higher than that of procaspase-8. This allows quick recruitment of c-FLIP to the DISC upon threshold stimulation in accordance with model predictions.
Stimulation with low concentrations of anti-APO-1 antibodies can result in two scenarios. First, upon stimulation at the threshold concentration of anti-APO-1 caspase activity would be observed, however at a level low enough to not result in apoptosis. According to the second scenario caspase activation does not take place upon stimulation with threshold concentrations of anti-APO-1. The first scenario contradicts our mathematical modeling (1) as well as the experimental data (Fig. 1). The second scenario involves inhibition of caspase-8 activation at the DISC level. Furthermore, the second scenario was predicted by the model and confirmed by the original set of experimental data (1). The data presented in this study further confirm the model predictions, demonstrating in detail inhibition of caspase-8 activation at the DISC that consequently prevents the following apoptosis signaling cascade.
To establish the functional relevance of stimulation with low concentrations of a CD95 stimulus we analyzed survival signaling, e.g. MAP kinase activation. Interestingly, we demonstrated that MAP kinase activation is not stopped upon threshold stimulation when death is blocked. These results go along with the fact that MAP kinase activation is not blocked by the addition of the pan-caspase inhibitor Z-VAD-fmk. This indicates that the mechanism of survival signaling via CD95 does not involve death components of the CD95 signaling pathway and caspase-8 activation. Transduction of the survival signal might occur via as yet uncharacterized proteins associated to the CD95. Our findings go along with the reports of survival signaling from CD95 in other cell types other than lymphocytes, e.g. primary neurons (15). In these cells the concentration of "death components" associated to the CD95 receptor might be lower than "survival components", resulting in the preferential signaling of survival. Interestingly, our study also goes along with the findings that cells carrying heterozygous mutations in the CD95 DD are deficient in induction of apoptosis but can still efficiently activate the transcription factor NF-B and the MAP kinases ERK1/2 and p38 (23).
Based on our studies we suggest the following model for CD95 signaling in CD95-sensitive cell lines (Fig. 5). At high doses of CD95 stimuli the CD95 DISC is formed and caspase-8 is activated, triggering the apoptotic cascade (Fig. 5, left). Survival signaling takes place in parallel; however, apparently it is not strong enough to prevent the cells from dying. At threshold concentrations c-FLIP molecules block caspase-8 activation at the DISC in accordance with the above described mechanism and survival signaling becomes the major event (Fig. 5, right).
Understanding of the switch mechanisms between life and death of cells plays an important role in therapeutic interven-tion strategies in diseases that are associated with apoptosis. Thus, the detailed molecular mechanism of signal transduction for survival signaling from CD95 needs to be addressed further in future studies. It should also be noted that this study once again shows that mathematical modeling in combination with biochemical analysis is a powerful tool that allows a detailed characterization of complex signaling pathways.