Phosphatidylinositol 3′-Kinase Blocks CD95 Aggregation and Caspase-8 Cleavage at the Death-Inducing Signaling Complex by Modulating Lateral Diffusion of CD95¹

Apoptotic death of lymphocytes is an important regulator of many steps in the function of immune responses. Differences in the ability of lymphocytes to respond to immune stimuli can lead to divergence in the apoptotic pathway. For example, mature Th1 cells are sensitive to CD95-mediated apoptosis upon CD3 ligation (1, 2). This apoptosis is mediated by CD95 (Fas/APO-1), a member of the death receptor subfamily of TNF/nerve growth factor receptors (2, 3, 4). The signal transduction pathway to apoptosis upon ligation of CD95 includes recruitment of the adaptor molecule Fas-associated death domain protein and procaspase-8 to the cytoplasmic portion of CD95 to form a death-inducing signaling complex (DISC)³ (3, 4). At the DISC caspase-8 is autoproteolytically cleaved in a two-step process that releases catalytically active caspase fragments (p18 and p10) into the cytosol. In contrast, when CD3 is ligated on Th2 cells, they up-regulate phosphatidylinositol 3′-kinase (PI 3′-K) and become resistant to CD95-mediated apoptosis (2). Up-regulation of PI 3′-K prevents generation of active caspase-8 subunits through the premature termination of the DISC-associated cleavage process (2).

The critical need for CD95 aggregation to initiate signaling for apoptosis is evidenced by several studies. Soluble CD95 ligand that can only trimerize CD95 (5, 6) fails to process caspase-8 (7). However, cross-linking of receptor or pro-caspase fusion proteins induces death (8, 9), demonstrating that multimerizing CD95is necessary to potentiate the inherent proteolytic activity of caspase-8 to transactivate a closely neighboring caspase. Earlier results comparing the aggregating IgG3 CD95 agonistic anti-APO-1 Ab with isotype-switched nonaggregating Abs also allude to the possibility that receptor oligomerization is necessary to activate caspase-8 (10). Based on this, we asked whether PI 3′-K mediated its anti-apoptotic effects on T cells by preventing CD95 aggregation. Indeed, we show here that PI 3′-K inhibits caspase-8 cleavage by inhibiting aggregation of CD95. This inhibition is effected by an actin-dependent reduction in the lateral diffusion of CD95.

Tetanus toxoid-reactive Th2 clones and purified protein derivative-reactive Th1 clones were established by limiting dilution technique as previously described (11). Subset specificity of the clones was determined by measuring production of IFN-γ and IL-4. For each T cell clone, 1 × 10⁵ cells were added in 1 ml of Iscove’s modified DMEM and 10% human AB serum in 24-well tissue culture plates precoated with 2.5 μg of Ab to CD3 (BioSource International, Camarillo, CA). Supernatants from the stimulated cells were harvested 18–20 h later, and cytokine quantities were measured by sandwich enzyme immunoassay using appropriate pairs of capture Abs and biotinylated detecting Abs (BD Pharmingen, San Diego, CA). To expand cell numbers, clones were maintained in IL-2, with bi-weekly stimulation of Ag and autologous APCs. Cytokine profile of the clones was routinely reconfirmed after several rounds of stimulation. We have observed that even after several in vitro stimulations and expansions, the cytokine profile and Ag reactivity of the clones remain unchanged.

The anti-APO-1 monoclonal (IgG3, κ) is an agonistic Ab recognizing an extracellular epitope of CD95 (12). The anti-caspase-8 Abs used recognize two different functional domains of caspase-8. The N2 Ab recognizes an epitope adjoining the tandem death effector domain containing prodomain. The C15 Ab detects the catalytically active p18 large subunit (13). Biotinylated and FITC-conjugated DX11 anti-CD95 mAb was obtained from BD Pharmingen. Caspase-8 encoding plasmid was a gift from Vishwa Dixit (Genentech, San Francisco, CA). In vitro transcription and translation was performed using the TnT Quick Coupled System obtained from Promega (Madison, WI). Wortmannin and cytochalasin D were purchased from Sigma (St. Louis, MO) and used at a final concentration of 1 μM.

Association and cleavage status of caspase-8 at the DISC was determined in 10⁷ T cells obtained after varying treatment conditions. Following the varying treatment conditions, T cells were harvested, washed in cold buffer, pelleted, and lysed in ice-cold lysis buffer (as described in Refs. 2, 3) for 30 min. Cell debris was pelleted by centrifugation for 10 min at 13,000 × g. Lysates were supplemented with 50 μl of rabbit anti-mouse (Cappel, Cochranville, PA)-coated staphylococcus A (Sigma). DISCs were allowed to immunoprecipitate in the cold for at least 1 h, then were washed in cold TNE (1 M Tris (pH 8), 1.5 M NaCl, 0.5 M EDTA, 1% Nonidet P-40) buffer. Immune complexes and staphylococcus A were separated after six washes by 15 min incubation in 2-ME containing sample buffer, staphylococcus A was spun out, and supernatants containing the immunoprecipitates were boiled. For Western blotting, lysates were separated on 12% SDS-PAGE, transferred to supported nitrocellulose membrane (Bio-Rad, Hercules, CA), and blocked with nonfat milk-containing Tween 20 + TBS (TTBS). Blots were incubated with primary Ab diluted in TTBS overnight at 4°C, washed three to four times for 10 min each with TTBS, then detected with HRP-conjugated secondary Ab, and developed using the enhanced chemiluminescence method (Amersham, Buckinghamshire, U.K.) following the manufacturer’s protocol. Multimeric status of CD95 was determined by fractionating either immunoprecipitated DISC or whole cell lysates under nonreducing conditions on a 5–15% gradient gel (Jule, New Haven, CT), Western blotted using the biotinylated DX11 mAb, and visualized with streptavidin-coupled HRP and ECL substrate.

In vitro transcribed/translated ³⁵S-methionine-labeled caspase-8 was incubated with immunoprecipitated DISC obtained from 30 × 10⁶ T cell clones resuspended in 50 μl cleavage assay buffer (50 mM HEPES, pH 7.4, 100 mM NaCl, 0.1% 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS), 10 mM DTT, 20% sucrose) (13) for 24 h at 4°C. Following incubation, pellets and supernatants were separated by centrifugation. Pellets were washed five times in DISC lysis buffer, resuspended in reducing sample buffer, and applied to a 12% SDS-PAGE gel. Supernatants were similarly subjected to SDS-PAGE separation. Gels were dried and exposed to film. Alternatively, gels were transferred to nitrocellulose and probed with caspase-8-specific Abs. Quantitative analysis of in vitro cleavage was determined by incubating cleavage assay supernatant with the caspase-8 colorimetric substrate Ile-Glu-Thr-Asp-p-nitroaniline (IETD-pNA) (BioVision, Palo Alto, CA) per manufacturer instructions.

Lateral diffusion was measured using the technique of FRAP (14). CD95 on T cell clones was fluorescently labeled with the nonagonistic DX11 Ab. Cells were washed twice before being loaded into 0.05 mm path length glass capillaries (Vitro Dynamics, Rockaway, NJ) and mounted onto glass slides with nail polish. Measurements of lateral diffusion and mobility were taken on a Zeiss Axioplan fluorescence microscope using a 1.30 63X NA Zeiss Plan Neo-fluor objective to focus a 488 nm argon ion laser. Data were collected and analyzed using custom software (14). Diffusion coefficient is expressed as 10⁻¹⁰ cm²/s.

We first compared DISCs immunoprecipitated from extracts of Th2 cells on the path to apoptosis after cross-linking of CD95 with DISCs from Th2 cells that were protected from apoptosis by concomitant ligation of CD3. Immunoprecipitated DISCs were separated under nonreducing conditions on a gradient gel to visualize receptor aggregation (Fig. 1,A). SDS-stable high molecular mass tetrameric aggregates of CD95 were isolated from cells whose CD95 was ligated with anti-APO-1 Abs (Fig. 1,A, lane 1). In contrast, cells treated with anti-APO-1 and anti-CD3 did not yield SDS- stable CD95 aggregates (Fig. 1,A, lane 2), but only receptor dimers. When PI 3′-K activity was inhibited with wortmannin, tetrameric CD95 aggregates were isolated from cells treated with both anti-CD3 and anti-APO-1, indicating that conditions that induce apoptosis lead to formation of higher order receptor oligomers (Fig. 1 A, lane 3). The CD95 aggregates were ∼200 kDa in size, suggesting that formation of an active complex required a minimum of four CD95 molecules to aggregate in the membrane. This aggregation is ligand induced because only monomeric CD95 was isolated from control unstimulated cells (data not shown). Preassociated trimers (15) were not observed, as no chemical cross-linking was performed to preserve the trimeric structures in the absence of ligand.

DISCs isolated from anti-APO-1-treated cells recruited and efficiently cleaved exogenous caspase-8 producing the p26 prodomain (Fig. 1,B, lane 1). In contrast, DISCs isolated from anti-CD3- and anti-APO-1-stimulated cells only processed exogenous caspase-8 to the p41/43 fragment, and failed to produce the p26 subunit (Fig. 1,B, lane 2). Cells treated with wortmannin before ligation of CD3 and CD95 were not restricted in their ability to process exogenous caspase-8, and generated the p26 prodomain fragment (Fig. 1,B, lane 3). As reported previously, we failed to immunoprecipitate DISCs from cells that received only CD3 stimulation (data not shown). Complete cleavage of exogenous caspase-8 was confirmed by examining the supernatants of the in vitro cleavage assay for the presence of catalytically active p18 subunit. Only under conditions where p26 fragment was generated was the presence of p18 also observed (Fig. 1,C, lanes 1 and 3). Conditions that lead to an active DISC also led to a 6-fold increase in caspase-8 activity measured by cleavage of IETD-pNA, whereas conditions where cells cannot generate active DISC and are protected from death failed to cleave IETD-pNA (Fig. 1 D). Thus DISCs from cells that are susceptible to CD95 triggered death, and can fully process caspase-8, consist of SDS-stable aggregates of CD95, whereas DISCS from cells protected from apoptosis by activation of PI 3′-K lack the CD95 aggregates necessary for full DISC activity and complete caspase-8 cleavage.

The lack of CD95 aggregates in CD3-stimulated cells raised the possibility that PI 3′-K inhibits lateral diffusion of CD95 and so reduces the probability of their aggregation and formation of functional DISCs. Therefore, we compared the lateral diffusion of CD95 in untreated and CD3-stimulated cells using FRAP. CD95 receptors of three individual T cell clones were fluorescently labeled with the nonagonistic CD95 Ab, DX11. A mobile fraction (40–50% of the total) of CD95 molecules diffused in the membrane of unstimulated cells with a diffusion coefficient, D∼30 × 10⁻¹⁰ cm²/s (Fig. 2, left panel). In contrast, after CD3 ligation, D of CD95 from a parallel set of T cell clones was 1 order of magnitude smaller, 3 × 10⁻¹⁰ cm²/s (Fig. 2, middle panel); the mobile fraction of receptors remained unchanged (data not shown). The decrease in D for CD95 following CD3 ligation, which correlated with the inability of CD3-stimulated cells to rapidly form CD95 aggregates, suggested that PI 3′-K plays a role in regulating the lateral diffusion of CD95 movement in the plasma membrane. Consistent with this idea, inhibition of CD3-triggered PI 3′-K activity by wortmannin returned D to values observed for unstimulated, apoptosis-sensitive cells (Fig. 2, right panel).

Previously we had reported that apoptosis-sensitive Th1 clones do not generate PI 3′-K activity following CD3 ligation and consequently caspase-8 is completely cleaved at the DISC in these cells (2). Therefore, to further confirm the specificity of PI 3′-K-mediated change in lateral diffusion, we examined CD95 lateral diffusion in a CD95-sensitive Th1 clone that does not generate PI 3′-K following CD3 stimulation. CD95 molecules diffused in the membrane of unstimulated cells with a diffusion coefficient, D ∼60 × 10⁻¹⁰ cm²/s (Fig. 2,B), comparable with unstimulated Th2 clones. However, in contrast to what we had observed in Th2 clones, CD3 stimulation did not alter CD95 lateral diffusion in Th1 cells (Fig. 2 B).

The role of PI 3′-K in regulating cytoskeletal networks is well documented (16, 17, 18, 19, 20, 21). The 10-fold reduction in D after CD3-stimulated PI 3′-K activation is consistent with steric hindrance of receptor movement by the cortical actin cytoskeleton (22, 23). Because the defect in caspase-8 processing correlates with hindered diffusion of CD95 to move in the membrane, we investigated the possibility that the cytoskeleton may be responsible for regulating CD95 movement.

To establish the linkage between PI 3′-K regulated CD95 aggregation and the cytoskeleton, we determined whether apoptosis, CD95 aggregation, caspase-8 cleavage, and CD95 lateral diffusion were all affected by cytochalasin D, which disrupts actin filaments. Cells that were treated with cytochalasin D, before CD95 ligation, were no longer protected from apoptotic death by CD3 ligation (Fig. 3). Lysates of these cells also yielded SDS-stable aggregates of CD95 (Fig. 4,A, lane 3), like untreated CD95-ligated cells (Fig. 4,A, lane 1). In contrast, cells that were CD3-stimulated before anti-APO-1 treatment, but not treated with cytochalasin D and were protected from apoptosis, did not yield CD95 aggregates (Fig. 4,A, lane 2). DISCs from CD3-activated T cells, treated with cytochalasin D just before CD95 ligation, fully processed caspase-8 (Fig. 4,B, lane 3) just as well as unactivated CD95-ligated cells (Fig. 4,B, lane 1) as evidenced by the presence of the p26 prodomain at the DISC, whereas CD3 activation without cytochalasin D treatment before CD95 ligation led to incomplete cleavage and an absence of p26 prodomain fragment at the DISC (Fig. 4,B, lane 2). Cytochalasin D treatment of activated cells also returned the diffusion coefficient of CD95 to that seen in untreated cells (Fig. 5) or of wortmannin-treated CD3-stimulated cells, ∼30 × 10⁻¹⁰ cm²/s⁻¹ (Fig. 2).

Importantly, PI 3′-K activity was substantially enhanced in CD3-activated T cells even with cytochalasin treatment. There was increased PI 3′-K activity following CD3 stimulation, which was inhibited in the presence of the PI 3′-K inhibitor, wortmannin (Fig. 6). However, cytochalasin treatment following CD3 ligation did not abrogate PI 3′-K activity in the cells (Fig. 6). At present we are not sure regarding the apparent synergy in PI 3′-K activity when CD3 and cytochalasin are used. Following CD3 activation, signaling molecules accumulate in rafts, and it is possible that the addition of cytochalasin excludes inhibitory molecules such as Cbl-b (24). However, of significance to this study is the fact that despite the presence of active PI 3′-K, when cytoskeletal constraints are removed CD95 receptors can diffuse freely to form tetrameric aggregates that are capable of fully processing caspase-8 and sensitizing cells to apoptosis.

Phosphorylation and inactivation of cell death proteins (25, 26) and death effector caspases (27) has been previously implicated as a mechanism engaged by PI 3′-K to inhibit apoptosis. The findings presented here delineate a new mechanism for PI 3′-K in regulating activity of death effector proteases. We show that PI 3′-K through its effects upon the actin cytoskeleton regulates the lateral diffusion of CD95 and thereby controls the state of CD95 aggregation and subsequent caspase-8 cleavage to its catalytically active forms. When CD95 is ligated in CD3-activated T cells the lateral diffusion of CD95 in the membrane is physically impeded by the reorganized actin cytoskeletal network. This lowered diffusion restricts the receptors from forming tetramers and higher order oligomers necessary to activate caspase-8. Thus when CD95 is ligated in CD3-activated cells, caspase-8 is recruited to the DISC, but because aggregation allowing for tetramer formation is blocked the DISC remains inactive and incapable of fully processing caspase-8. Overall, the data indicate that following APO-1 ligation, sensitive cells can form tetrameric aggregates of CD95 and resistant cells form only receptor dimers. The size of aggregates generated in response to membrane-associated CD95 ligand remains to be determined.

Although slow receptor diffusion could, with time, result in tetrameric aggregate formation, a competing process such as receptor internalization (28) may remove nascent receptor aggregates before they reach the critical size. Alternatively, inactivation or dispersion of nascent DISCs could occur. In this regard, FLIP (cellular FLICE inhibitory protein), a catalytically inactive form of caspase-8, which is also competitively recruited to the DISC (1, 2), can inactivate the nonmultimerized DISC by initiating an anti-apoptotic signaling response from CD95 (29). In addition, a short splice variant of FLIP is up-regulated in activated T cells and is strongly recruited to the DISC, where it blocks caspase-8 cleavage (30). The exact stoichiometry of caspase-8 and FLIP molecules that get recruited to the DISC following CD95 receptor ligation is undetermined. Nevertheless, the fact that both FLIP and caspase-8 molecules are capable of getting recruited into the DISC necessitates rapid CD95 aggregate formation to allow adequate caspase-8 molecules to be present in the DISC for efficient trans-proteolytic activation. Hence the ability to rapidly form aggregates is critical in determining the outcome of CD95 ligation.

Diffusion coefficient of many membrane proteins appears to depend on their exodomains (31, 32, 33). Therefore, regulation of D, rather than mobile fraction by the cytoskeleton is unusual, and contrasts with the results on aggregation of Fc receptors (34, 35, 36) and CD2 (37), in which aggregated receptors interact with the cytoskeleton and are immobilized and internalized. Because we do not see changes in the mobile fraction of CD95 following CD3 ligation, it is unlikely that lateral diffusion of CD95 is impaired by direct tethering to the cytoskeleton. We instead propose that lateral diffusion of CD95 in the membrane is physically impeded due to corralling by reorganized actin cytoskeletal networks. Studies of the interactions between the erythrocyte membrane protein Band 3 and the erythrocyte membrane cytoskeleton support a paradigm very similar to the one we propose for CD95 (22, 23). In this model, it is suggested that the cytoplasmic portion of Band 3 collides with the mesh of the membrane cytoskeleton and becomes fenced within the actin cytoskeletal grid, resulting in slowed diffusion in the membrane (23). In a like manner, lateral diffusion of transferrin receptor molecules in the plasma membrane of cells are also regulated by the dynamically fluctuating membrane cytoskeleton (38).

A recent study by Siegel et al. (15) revealed that trimerization of CD95 is not ligand dependent, but preassociated trimers exist on the cell surface of T cells. Sixty percent of the CD95 chains exist in the preassociated trimeric state, but are unable to recruit caspase-8 and initiate signaling for apoptosis. Our results show that CD95 has to minimally tetramerize to recruit and cleave caspase-8 to its catalytically active form. Taken together, these data imply that preassociated trimers have to encounter another monomeric receptor to form active complexes in response to treatment with agonistic APO-1 Abs. Following CD3-stimulation, diffusion of both monomers and trimers is restricted, preventing the formation of active tetramers. At present it is not clear how dimers are formed in CD3-stimulated cells following APO-1 treatment. A possible explanation is that, despite the caging effect induced by the polymerized actin in CD3-stimulated cells, there is likely a power law relationship between frequency of correctly sized gaps in the cytoskeleton and size of the diffusing molecules that allows monomers to diffuse slightly faster than trimers.

In conclusion, the findings reported here describe a novel mechanism for how PI 3′-K regulates CD95 function by a dynamic membrane process. Unlike most receptors that can signal following either their dimerization or trimerization, we show that CD95 has to become aggregated into tetramers to initiate the apoptotic signaling pathway. The necessity for receptor aggregation favors modulation of lateral diffusion as a unique mechanism to regulate CD95 function.