Inhibition of phosphatidylinositol 3-kinase sensitizes vascular endothelial cells to cytokine-initiated cathepsin-dependent apoptosis.

In the presence of cycloheximide, tumor necrosis factor or interleukin-1 initiates caspase activation, loss of mitochondrial membrane potential (DeltaPsi), DNA degradation, and nuclear condensation and fragmentation characteristic of apoptotic cell death in human vascular endothelial cells (EC). Inhibition of phosphatidylinositol 3-kinase (PI3K) by LY294002, but not inhibition of Akt by dominant-negative mutation, also sensitizes EC to cytokine-initiated apoptosis. Cytokine-initiated caspase activation is slower and comparatively less with LY294002 than with cycloheximide. Cycloheximide but not LY294002 decreases expression of c-FLIP (cellular FLICE inhibitory protein), an inhibitor of caspase-8 activation. The caspase inhibitor zVADfmk completely blocks caspase activation, DNA degradation, and nuclear fragmentation in both cases but only prevents loss of DeltaPsi and cell death for cytokine plus cycloheximide treatment. In contrast, overexpression of Bcl-2 protects EC treated with cytokine plus LY294002 but not EC treated with cytokine plus cycloheximide. The cathepsin B inhibitor CA-074-Me prevents loss of DeltaPsi, caspase activation, and cell death for EC treated with cytokine plus LY294002 but has no effect on EC treated with cytokine plus cycloheximide. Cathepsin B translocates from lysosomes to cytosol following treatment with LY294002 prior to the activation of caspases. These results suggest that inhibition of PI3K allows cytokines to activate a cathepsin-dependent, mitochondrial death pathway in which caspase activation is secondary, is not inhibited by c-FLIP, and is not essential for cell death.

In the presence of cycloheximide, tumor necrosis factor or interleukin-1 initiates caspase activation, loss of mitochondrial membrane potential (⌬⌿), DNA degradation, and nuclear condensation and fragmentation characteristic of apoptotic cell death in human vascular endothelial cells (EC). Inhibition of phosphatidylinositol 3-kinase (PI3K) by LY294002, but not inhibition of Akt by dominant-negative mutation, also sensitizes EC to cytokine-initiated apoptosis. Cytokine-initiated caspase activation is slower and comparatively less with LY294002 than with cycloheximide. Cycloheximide but not LY294002 decreases expression of c-FLIP (cellular FLICE inhibitory protein), an inhibitor of caspase-8 activation. The caspase inhibitor zVADfmk completely blocks caspase activation, DNA degradation, and nuclear fragmentation in both cases but only prevents loss of ⌬⌿ and cell death for cytokine plus cycloheximide treatment. In contrast, overexpression of Bcl-2 protects EC treated with cytokine plus LY294002 but not EC treated with cytokine plus cycloheximide. The cathepsin B inhibitor CA-074-Me prevents loss of ⌬⌿, caspase activation, and cell death for EC treated with cytokine plus LY294002 but has no effect on EC treated with cytokine plus cycloheximide. Cathepsin B translocates from lysosomes to cytosol following treatment with LY294002 prior to the activation of caspases. These results suggest that inhibition of PI3K allows cytokines to activate a cathepsin-dependent, mitochondrial death pathway in which caspase activation is secondary, is not inhibited by c-FLIP, and is not essential for cell death.
Vascular endothelial cells (EC) 1 are a principal target of the pro-inflammatory cytokines TNF and IL-1. Effects on the en-dothelium that contribute to the inflammatory response are largely dependent on gene transcription resulting in the expression of proteins controlling vasoregulation, leukocyte adhesion, leukocyte activation, and coagulation. In some cases TNF and IL-1 may also result in endothelial injury, a common feature in the pathogenesis of vascular leak, sepsis, and transplant rejection. The ability of TNF or IL-1 to cause endothelial injury may occur indirectly through the activation and recruitment of leukocytes or generation of thrombosis or may occur directly from the pro-apoptotic actions of these cytokines on EC.
The direct pro-apoptotic action of TNF on various cell types generally results from the ligand-dependent assembly of a death-inducing signaling complex (DISC), so called for the ability of this complex to initiate caspase activation. For TNF, the formation of a DISC is dependent on ligand-binding to TNF receptor type I (TNFR-1, also designated CD120a), which leads to the recruitment of the cytosolic adapter protein TNFR-1 associated death domain protein (TRADD) (1). The association between TNFR1 and TRADD involves the "death domains" (DD) of these proteins. DDs are homologous regions of ϳ80 amino acids that mediate protein-protein interaction and are also found in other receptors such as Fas (CD95) (2). TRADD may subsequently recruit Fas-associated death domain protein (FADD) through DD interactions (3). FADD contains both a DD and a death effector domain, the latter of which can interact with either pro-caspase-8 (also known as FLICE) or with cellular FLICE inhibitory protein (c-FLIP) (4,5). FADD-associated pro-caspase-8 undergoes autocatalytic activation by proteolysis, liberating the active enzyme from the pro-form (6). c-FLIP may inhibit this process.
Activated caspase-8 dissociates from the DISC and acts on various cytosolic substrates. For example, caspase-8 may proteolytically activate the effector caspase-3. Activated caspase-3, in turn, cleaves a variety of substrates, resulting in apoptotic cell death (7). Alternatively, caspase-8 may proteolytically activate a cytosolic protein called Bid (8). Bid is a pro-apoptotic Bcl-2 family member containing a single Bcl-2 homology (BH) domain designated BH3. Proteolytically activated forms of "BH3-only" family members, such as Bid and Bad, bind to mitochondrial-associated "BH1-3" or "multidomain" proteins, such as Bax and Bak, causing supramolecular openings of the outer mitochondrial membrane (9 -12). These openings allow the release of cytochrome c from the mitochondria (13). Within the cytosol, cytochrome c associates with Apaf-1, and this complex further recruits pro-caspase-9 forming an assembly sometimes called an apoptosome. The apoptosome mediates ATPdependent autocatalytic processing of caspase-9 and activated caspase-9, like caspase-8, can catalyze the proteolytic activation of caspase-3 resulting in apoptotic cell death (14). Thus caspase-8-mediated cleavage of Bid activates an amplification pathway for mitochondrial-dependent activation of caspase-3. Mitochondrial openings may also release other proteins, such as apoptosis-inducing factor that can initiate caspase-independent cell death (15). The requirement for the Bid/cytochrome c/Apaf1 amplification pathway differs among various cell types and correlates with the extent to which active caspase-8 is generated by the DISC (8,16). DISC activity may be positively regulated by the expression levels of FADD and pro-caspase-8 (17) or negatively regulated by the levels of c-FLIP (5).
IL-1, like TNF, initiates the activation of signal transduction cascades by the recruitment of adapter proteins to its receptor. In EC, IL-1 may also initiate apoptotic cell death. To date, the precise components of the IL-1 receptor-associated DISC have not been defined. A number of adapter proteins involved in IL-1 signal transduction, namely MyD88 and IRAK (18), contain DDs and are possible mediators of caspase recruitment and activation.
Most untransformed cell types are not sensitive to the proapoptotic actions of TNF or IL-1, unless mRNA translation or protein synthesis is blocked. This observation has been explained by the capacity of TNF to stimulate the activation of NFB, resulting in the up-regulation of anti-apoptotic gene products such as c-FLIP, XIAP, c-IAP 1, and c-IAP 2 (19 -21).
FIG. 1. Inhibition of protein synthesis or PI3K sensitizes EC to TNF-or IL-1-mediated cell death. EC were stimulated with TNF or IL-1 (10 ng/ml) either alone or in the presence of cycloheximide for 8 h (top panels), or EC were pretreated with vehicle with or without LY294002 (50 M) for 3 h prior to the addition of TNF or IL-1 (10 ng/ml) and continued incubation overnight (bottom panels). For both treatments, floating and attached EC were harvested, pooled, cytospun at 1000 rpm for 3 min, stained with DAPI, and visualized by fluorescence microscopy.
The expression levels of several anti-apoptotic genes, such as c-FLIP and IAP 1 are also regulated by the proteasome (22,23). In the presence of cycloheximide (CHX), levels of c-FLIP are rapidly diminished, favoring DISC-dependent activation of caspase-8 (19,22). Reduction of c-FLIP by antisense oligonucleotides mimics the effect of CHX and similarly sensitizes cells to death (22).
Although apoptosis is generally associated with caspase activation, either through a DISC or through an apoptosome, caspase-independent cell death has been observed with the generation of apoptotic-like features in a variety of cell types (24 -26). In many instances, these variant forms of apoptosis are mediated by non-caspase proteases such as the cysteine protease cathepsin family, calpains, serine proteases, or the proteasome complex. Of particular interest, the activation of cathepsin B has been shown to play a central role in the generation on TNF-mediated cell death in fibrosarcoma cells (25). Furthermore, TNF-mediated apoptosis has been shown to be strongly reduced in hepatocytes from cathepsin B-deficient mice (27). This pathway has not been described in a normal (untransformed) human cell type.
While most TNF activities on EC result in inflammation and/or apoptosis, we have shown in EC that TNF and IL-1 also activate the anti-apoptotic phosphatidylinositol 3-kinase/Akt pathway (28). PI3K converts plasma membrane phosphatidylinositol 4,5-bisphosphate into phosphatidylinositol 3,4,5trisphosphate, catalyzing the recruitment of several enzymes, such as PI3K-dependent protein kinase and Akt to the plasma membrane. Akt is a serine/threonine kinase that exerts an anti-apoptotic action by the phosphorylation of a number of substrates containing the phosphorylation consensus RXRXX(S/T). Akt-mediated phosphorylation inactivates the pro-apoptotic Bcl-2 homologue Bad (29), the apoptosis-initiating enzyme caspase-9 (30), and the forkhead family transcription factor FKHRL1 (31), which mediates transcription of proapoptotic gene products. In some cells Akt may regulate the activity of NFB either through direct phosphorylation and activation of IKK␣ (32) or through regulation of the transactivation capacity of Rel A (33). We have shown that activation of PI3K in EC does not contribute to the activation of NFB or have any significant effect on NFB-dependent inflammatory responses (28). Similarly, Akt is reported not to have any effect of NFB activation in HeLa cells (34).
In the present study we have examined the role of PI3K and Akt activation in the regulation of apoptosis induced by TNF and IL-1. To do so, we either inhibited PI3K with LY294002 or inhibited Akt by retroviral transduction with an inactive (K179M), dominant-negative form of Akt. We report that inhibition of PI3K but not of Akt sensitizes EC to the apoptotic actions of TNF and IL-1 and that the cell death caused by this pathway could not be blocked by caspase inhibition with zVADfmk but instead appears to be mediated through cathepsin B.

EXPERIMENTAL PROCEDURES
Cell Culture-Human umbilical vein EC were isolated from discarded tissue in accordance with an approved protocol by Yale University Human Investigations Committee and serially cultured on gelatin (J. T. Baker Inc., Phillipsburg, NJ)-coated tissue culture plastic (Falcon, Lincoln Park, NJ) in Medium 199 (M199) supplemented with 20% fetal calf serum, 200 M L-glutamine (all from Invitrogen, Grand Island, NY), 50 g/ml EC growth factor (ECGF) (Collaborative Biomedical Products, Bedford, MA), 100 g/ml porcine heparin (Sigma Chemical Co., St. Louis, MO), 100 units/ml penicillin, and 100 g/ml streptomycin (Invitrogen). All experiments were performed using EC at passage 2 or 3. Such sub-cultured EC uniformly display CD31 and von Willebrand factor and are free of CD45 ϩ -contaminating leukocytes. For replating assays floating EC were harvested, washed with Hanks' balance salt solution, and re-seeded onto gelatin-coated plates. EC that remained substrate-attached were similarly re-plated following harvesting with trypsin (Invitrogen).
Immunoblotting-For immunoblots, each well of a six-well plate containing a confluent HUVEC monolayer was washed twice in ice-cold PBS and lysed by the addition of 100 l of lysis buffer (50 mM Tris-Cl, pH 6.8, 150 mM NaCl, and 1% Triton X-100) supplemented with Pefabloc (1 mM) and complete protease inhibitor mixture. For the measurement of phospho-proteins, NaF (10 mM) and Na 3 VO 4 (1 mM) were also included in the lysis buffer to reduce phosphatase activity. After 20 min on ice, lysates were harvested by scraping. Where indicated, detached EC were harvested by centrifugation, washed in PBS, and pooled with the lysate of the attached EC from the same sample. For each sample, an equal amount of protein was separated by SDS-PAGE (35) then transferred electrophoretically to a polyvinylidene difluoride membrane

FIG. 2. Inhibition of Akt does not account for the sensitization to cytokine-mediated cell death following inhibition of PI3K.
LZRS or LZRS-K179M Akt transduced EC were stimulated with TNF (10 ng/ml) for 0 -45 min. Lysates were separated by SDS-PAGE, transferred to polyvinylidene difluoride membranes, and immunoblotted for HA, FKHR, or phospho-FKHR (A). LZRS or LZRS-K179M Akt transduced EC were treated overnight with TNF or IL-1 (10 ng/ml) either alone, following pretreatment with LY294002 (50 M), or in combination with CHX. Floating and attached EC were harvested, pooled, EtOH-fixed, stained with propidium iodide (50 g/ml), and analyzed for hypodiploid DNA by FACS (B).
(Immobilon P, Millipore, Milford, MA) and immunoblotted with primary and horseradish peroxidase-conjugated secondary antibodies. Detection of the bound antibody by enhanced chemiluminescence was performed according to the manufacturer's instructions (Pierce Chemical Co., Rockford, IL).
Cell Cycle and Hypodiploid DNA Analysis-EC grown to confluence on 12-well plates were treated as described in the text. At the indicated time after treatment, floating EC were collected and pooled with residual attached EC suspended by trypsin treatment. The pooled EC were washed once in PBS and fixed by resuspension in 70% ethanol for 15 min. After fixation, EC were washed once more in PBS before incubation in PBS containing propidium iodide (50 g/ml) and RNase A (1 mg/ml) for 0.25-2 h. The DNA content of EC was then determined by FACS analysis using Cell Quest software (FACSort, BD Biosciences, San Jose, CA).
DAPI Staining of EC-EC were grown to confluence on gelatincoated 12-well plates and treated as described in the text. After treat-ment, attached EC were harvested with trypsin and combined with floating EC harvested from the same sample. Cells were washed in PBS, and ϳ1 ϫ 10 5 cells were adhered to a glass coverslip by spinning in a cyto-centrifuge (Shandon, Pittsburgh, PA). Slides were air-dried and dipped in a chamber containing MeOH and DAPI (1 g/ml). After rinsing in PBS, a drop of Gel Mount (Biomeda Corp., Foster City, CA) and a coverslip was placed over the cells. Specimens were examined by immunofluorescence microscopy using a Nikon diaphot microscope with a 360-nm filter.
Quantitation of EC Adherence and Replating-To quantify the number of cells that remained adherent, EC plated on gelatin-coated 96-well plates were treated as indicated in the text. After experimental manipulation the medium was removed and cells were washed twice in PBS. The residual attached cells were fixed and stained by the addition of 70% ethanol containing 100 g/ml Hoescht 33258 reagent (Molecular Probes, Eugene, OR) for 30 min at room temperature. Cells were again washed twice with PBS, and the residual fluorescence was re- corded ( ex ϭ 360 nm, em ϭ 460 nm) using a fluorescence plate reader (Perspective Biosystems Inc, Framingham, MA). To assess the viability of detached versus adherent cells, EC were grown to confluency in 12-well plates and treated as indicated in the text. Detached EC were harvested, washed in Hanks' balanced salt solution, and re-seeded onto gelatin-coated plates. EC that remained substrate-attached were similarly re-plated following harvest with trypsin. Viability was assessed as replating efficiency 18 h later quantified by Hoescht staining as above.
Caspase Activity Assays-For experimental manipulation EC were plated on 12-well plates and treated at confluency as indicated in the text. After described treatment the CaspaTag peptide (FAM-VAD-fmk for broad range caspase activity, FAM-LETD-fmk for caspase-8, FAM-LEHD-fmk for caspase-9, or FAM-DEVD-fmk for caspase-3) was added to each well and incubated a further 1 h according to the manufacturer's instructions. Subsequent to incubation with the peptide, floating EC were harvested and combined with attached EC from the same well harvested with trypsin. Caspase activity was demonstrated by the generation of a second peak or shoulder on FL-1 that results from peptide binding to active caspase by FACS.
Mitochondrial Membrane Potential (⌬⌿) Analysis-After experimental manipulation of EC seeded on 12-well plates, floating EC were harvested by centrifugation and combined with remaining substrateattached EC harvested with trypsin. The pooled EC were washed 1ϫ in PBS containing 1% bovine serum albumin before resuspension in 200 l of PBS/bovine serum albumin containing JC-1 (10 g/ml). After 15 min of incubation at 37°C, EC were washed, re-suspended in PBS, and analyzed by FACS.
Retroviral Transduction-HA-tagged murine K179M Akt (a gift from Dr. W. Sessa, Yale University) was Topo-cloned using EcoRI and NotI into the retroviral LZRS expression vector (a gift from Dr. A. L. M. Bothwell, Yale University), and the construct was verified by sequencing. The caspase-resistant Bcl-2 retroviral construct (a gift from A. L. M. Bothwell) has been described previously (36). The amphotropic Phoenix packaging cell line was transfected with either the empty vector LZRS, LZRS-K179M Akt, or LZRS-Bcl2 using LipofectAMINE (Invitrogen) and selected for gene expression 24 h after transfection using puromycin (1 g/ml). Puromycin-resistant cells were used to derive conditioned medium to provide a retroviral stock for HUVEC transduction. For transduction of primary HUVECs, M199 containing ECGF was removed, and cells were washed and incubated 5-8 h with retroviral conditioned media containing Polybrene (8 g/ml, Sigma). After incubation, retrovirus was removed and replaced with normal growth medium overnight. The transduction process was repeated a further three times with intermittent cell passage as required. Using this protocol the percentage of HUVECs expressing the transgene is routinely Ͼ95%.
Preparation of Cytosolic Extracts for the Analysis of Cathepsin B-Measurement of cytosolic cathepsin B was determined using methodology similar to Foghsgaard et al. (25). Endothelial cells were pretreated with LY294002 (50 M) for 3 h in complete M199 prior to the addition of cytokine for a further 3 h. After treatment media were removed and cells were washed twice in PBS prior to the addition of extraction buffer (50 g/ml digitonin, 250 mM sucrose, 20 mM Hepes, 10 mM KCl, 1.5 mM MgCl 2 , 1 mM EDTA, 1 mM EGTA, 1 mM Pefabloc, pH 7.5) and incubation on ice for 20 min. (The conditions that allow for the selective permeabilization of the plasma membrane by digitonin without perturbation of lysosomes were determined in preliminary experiments using EC pre-loaded both with calcein-AM and LysoTracker red.) After incubation, the cytosolic extract was collected. Samples were analyzed for cathepsin B either by Western blotting or with a cathepsin B activity assay.
Measurement of Cathepsin B Activity-A 50-l volume of cytosolic extract was added to an equal volume of cathepsin reaction buffer (50 mM sodium acetate, 4 mM EDTA, 8 mM dithiothreitol, 1 mM Pefabloc, pH 6.0). Cathepsin B activity was measured by the addition of 20 M Z-Arg-Arg-AMC (Calbiochem). Liberated AMC was measured ( ex ϭ 360 nm, em ϭ 460 nm) using a fluorescence plate reader immediately following the addition of the peptide substrate (T 0 ) and following a 60 min incubation at 37°C (T 60 ). Activity was determined by subtracting the background activity at T 0 from activity at T 60 and correcting for the amount of protein in each sample.

RESULTS
TNF and IL-1 Mediate Nuclear Condensation and EC Death under Conditions Where Either Protein Synthesis Is Inhibited or PI3K Activation Is Blocked-Many cell types are not sensi-tive to the pro-apoptotic actions of TNF or IL-1 unless RNA or protein synthesis is blocked. As previously described (37), treatment of EC with TNF or IL-1 (10 ng/ml) in the presence of CHX (2.5 g/ml) resulted in cell detachment (not shown) and in nuclear condensation and EC death (Fig. 1, top panels, E and  F). In the present investigation, we compared the effect of inhibition of PI3K using the pharmacological inhibitor LY294002 with the effects of inhibiting protein synthesis. At concentrations of LY294002 between 30 and 50 M, which we established to be effective in the inhibition of TNF-induced phosphorylation of Akt, we observed that pretreatment with LY294002 sensitized EC to the pro-apoptotic action of TNF or IL-1, resulting in extensive cell detachment (data not shown). EC that remained substrate-attached were able to re-adhere to gelatin-coated tissue culture plastic with close to 100% efficiency, whereas EC that had detached could not be re-plated, confirming that adherent cells remain viable, whereas detached cells are not. DAPI staining of cytospin preparations from pools of detached and viable cells confirmed that dead cells show nuclear condensation and fragmentation indicative of apoptosis (Fig. 1, bottom panels, E and F).
Inhibition of Akt Does Not Account for Sensitization to Cytokine-mediated Apoptosis-To determine whether the sensitization to TNF-or IL-1-mediated cell death upon inhibition of PI3K with LY294002 was dependent upon inhibition of Akt, EC were subjected to retroviral transduction with a dominantnegative inactive kinase form of Akt (K179M Akt). The functional expression of K179M Akt was confirmed by the inhibition of the TNF-dependent phosphorylation of the forkhead transcription factor (FKHR) (Fig. 2A). Surprisingly, expression of K179M Akt had no effect on cell survival following treatment with TNF or IL-1. Similarly, transduction with K179M Akt did not result in potentiation of cell death observed in response to LY294002 either alone or in combination with cytokine treatment. These experiments suggest that inhibition of Akt is not sufficient to explain the death response observed in response to cytokine following inhibition of PI3K. K179M Akt expression did cause a significant increase in the death response observed following treatment with cytokine plus CHX (Fig. 2B) suggesting that active Akt can limit caspase-induced cell death. The remainder of this study focused on the effects of LY294002 treatment. Activation of Caspase-8, -9, and -3 following Inhibition of PI3K or Treatment with Cycloheximide-The most proximal caspase activated by formation of a DISC is caspase-8. Activated caspase-8 may directly activate caspase-3 or, via its actions on the mitochondria, generate the apoptosome and activate caspase-9 (7,8). We analyzed the effects of cycloheximide and PI3K inhibition on TNF-or IL-1-mediated activation of caspase-8 and subsequent downstream caspases-9 and -3. Using a living cell, fluorometric caspase assay (38) we observed activation of caspases-8, -9, and -3 within 1 h of treatment with cytokine plus CHX, which increased to a maximal activity at around 6 h (Fig. 3A). We could not clearly discern by this method sequential activation of these caspases, although caspase-8 activation was in some experiments detected slightly earlier than activation of caspase-9 or -3 (not shown). Activation of caspases-8, -9, and -3 could also be observed in response to cytokine following inhibition of PI3K. In this case, however, caspase activation was slower and less extensive. No activation was observed for the first 6 h, and peak activity occurred only after an overnight incubation of 18 h or more (Fig. 3B).
Treatment of EC with Cycloheximide but Not LY294002 Decreases the Expression of c-FLIP-Sensitization of EC to cytokine-mediated apoptosis by TNF or IL-1 has been reported to result from the inhibition of NFB-dependent transcription of anti-apoptotic genes coupled with degradation of the apoptosis inhibitor protein c-FLIP (19,22). We confirm that treatment of EC with CHX results in a rapid decrease in the expression of c-FLIP (Fig. 4A). However, in contrast to the effect observed with CHX, inhibition of PI3K by treatment with LY294002 had no effect on the expression of c-FLIP, nor did it prevent TNF-or IL-1-induced up-regulation of c-FLIP (Fig. 4B).
Caspase Activation Is Not Required for Cytokine-initiated Cell Death following Inhibition of PI3K-Many of the features of programmed cell death such as detachment, DNA degrada- After treatment overnight (for TNF following LY294002) or for 6 h (TNF with or without CHX), the broad range caspase substrate FAM-VAD-fmk was added for 1 h further before EC were harvested and analyzed for caspase activity by FACS (A). EC were treated in the presence or absence of zVADfmk (25 M) with TNF or IL-1 (10 ng/ml) either alone, following pretreatment with LY294002 (50 M), or in combination with CHX (2.5 g/ml). After treatment overnight, detached EC were removed by washing, and substrateattached EC were quantified by Hoescht staining (B). EC were treated in the presence or absence of zVADfmk (25 M) with TNF or IL-1 (10 ng/ml) either alone, following pretreatment with LY294002 (50 M), or in combination with CHX (2.5 g/ ml). After treatment overnight with cytokine plus LY294002 or for 6 h with cytokine plus CHX, both floating and attached EC were harvested, EtOH-fixed, stained with PI (50 g/ml), and analyzed for hypodiploid DNA by FACS (C). tion, loss of mitochondrial membrane potential (⌬⌿), and nuclear condensation or fragmentation are dependent on the activation of effector caspases. To assess the role of caspase activation in the TNF-or IL-1-mediated death pathways activated following inhibition of PI3K or CHX treatment, we employed the broad range caspase inhibitor, zVADfmk. Titration of zVADfmk from 0 to 50 M resulted in a dose-dependent inhibition of caspase activity in response to TNF in the presence of either CHX or LY294002 (not shown) with full inhibition being achieved with zVADfmk at 25 M (Fig. 5A). This concentration was used in all further studies. Although zVADfmk effectively blocked the detachment of EC observed in response to TNF or IL-1 in the presence of CHX, the inhibitor had no effect or actually increased the detachment of EC in response to cytokine treatment following inhibition of PI3K (Fig.  5B). (Detached ECs were confirmed as being dead by the inability to re-attach to substrate following removal from the apoptotic stimulus (data not shown).) The disparity between the ability of zVADfmk to block cell death in response to cytokine plus CHX but not following inhibition of PI3K suggested the existence of a caspase-independent death pathway in the latter case. To further characterize the death pathway activated in response to TNF or IL-1 following inhibition of PI3K, we examined the effect of zVADfmk on some of the parameters associated with apoptosis, i.e. DNA degradation, nuclear condensation/fragmentation, and loss of ⌬⌿. Caspase inhibition blocked DNA degradation observed both in response to TNF or IL-1 following inhibition of PI3K or treatment with CHX assessed by PI staining and FACS analysis (Fig. 5C). Nuclear morphology by DAPI staining confirmed that treatment with TNF or IL-1 plus LY294002 resulted in the generation of fragmented nuclei (Fig. 1, bottom  panel, E and F). Caspase inhibition completely prevented nuclear condensation associated with TNF or IL-1 plus CHX (Fig.  1, top panels, H and I). However, in the presence of zVADfmk, EC were subjected to retroviral transduction with either empty vector LZRS or caspase-resistant Bcl-2. Stable transductants were treated with TNF or IL-1 (10 ng/ml) either alone, following pretreatment with LY294002 (50 M), or in combination with cycloheximide (2.5 g/ml). After treatment overnight, floating EC were removed by washing, and substrate-attached EC were quantified by Hoescht staining.
cytokine plus LY294002 still resulted in nuclear condensation but did not progress to fragmentation (Fig. 1, bottom panel, H and I). Treatment with TNF or IL-1 in the presence of CHX or following inhibition of PI3K resulted in loss of ⌬⌿ as observed by an increase in fluorescence (FL-1) with the dye JC-1 (Fig.  6A). Caspase inhibition with zVADfmk blocked loss of ⌬⌿ in response to cytokine plus CHX but had no effect on the loss of ⌬⌿ in response to cytokine following inhibition of PI3K (Fig.  6B). These data suggested that, in the presence of LY294002, TNF and IL-1 activated a mitochondrial death pathway that was not dependent upon caspase activation. However, in the absence of zVADfmk, caspase activation did occur and the activation of caspases altered the morphological and biochemical features of cell death.
Bcl2 Expression Blocks Cytokine-initiated Cell Death following Inhibition of PI3K but Not Protein Synthesis-To further examine the role of the mitochondria in the death pathways evoked by TNF or IL-1 following inhibition of PI3K or treat-ment with CHX, EC were subjected to retroviral transduction with a caspase-resistant form of Bcl-2. Four rounds of infection with either empty vector (LZRS) or LZRS-Bcl-2 resulted in over 95% expression of the transgene in EC as confirmed by intracellular FACS staining (described previously in Ref. 36). We observed that transduction with Bcl-2 protected EC from cytokine-initiated cell death and detachment following inhibition of PI3K but had no effect or actually increased the killing observed in the presence of CHX (Fig. 7). Similarly, Bcl-2 expression effectively blocked both the activation of caspase-8 and loss of ⌬⌿ observed following inhibition of PI3K but had no effect on the activation of caspase-8 or loss of ⌬⌿ in response to cytokine in the presence of CHX (Fig. 8, A and B). Together these results support the observation that caspase activation in response to cytokine treatment following inhibition of PI3K results from rather than causes alterations in the mitochondria. In contrast, cytokine-plus CHX-induced cell death occurs through a typical Type I death pathway that is completely FIG. 8. Bcl2 transduction protects EC from TNF-and IL-1-induced caspase-8 activation and loss of ⌬⌿ following inhibition of PI3K but not protein synthesis. LZRS (shaded histogram)-or Bcl2 (open histogram)-transduced EC were treated with TNF or IL-1 (10 ng/ml) either alone, following pretreatment with LY294002 (50 M) for 3 h, or in combination with CHX (2.5 g/ml). After incubation overnight the caspase-8 substrate FAM-LETD-fmk was added for 1 h further before floating, attached EC were harvested, and caspase activity was determined with FACS. The numbers in parentheses reflect the FL-1 mean for Bcl-2 Ϫ the FL-1 mean for LZRS (A). EC were treated as described for panel A before floating, and attached EC were harvested, washed, and resuspended in PBS containing JC-1 (10 g/ml). After 15 min at 37°C EC were washed again and ⌬⌿ was measured by FACS (B). caspase-dependent. Changes in the mitochondria of EC in this pathway occur downstream of caspase-8 activation and are not required for cell death.
Cathepsin B Activation Is Required for Cytokine-initiated Death following Inhibition of PI3K-We explored the possibility that the non-caspase protease cathepsin B could be the signal upstream of the mitochondria to initiate caspase-independent cell death following inhibition of PI3K using the specific cathepsin B inhibitor CA-074-Me. CA-074Me (30 M) effectively blocked all of the features of cell death in response to cytokine following inhibition of PI3K such as increased EC detachment, DNA fragmentation, caspase activity, and loss of ⌬⌿ (Figs. 9, A-C, and 10B). In contrast, inhibition of cathepsin B had little or no effect on the same parameters observed in response to cytokine in the presence of CHX (Figs. 9, A-C, and 10A). Direct evidence that the activation of cathepsin B occurred in response to cytokine treatment following inhibition of PI3K was provided by translocation of active cathepsin B to the cytosol. Following only 3 h of treatment with LY294002 with or without TNF (i.e. at an earlier time than we could detect caspase activation) we observed an increase in immunoreactive cathepsin B in the cytosol using Western blotting (Fig. 11A). The appearance of cytosolic cathepsin B was paralleled by an increase in the cytosolic activity of the enzyme that was completely blocked in the presence of CA-074-Me (30 mM) (Fig.  11B).
To identify the mechanism through which an increase in cathepsin B activity might effect mitochondria to result in a loss of ⌬⌿ and activation of downstream effector caspases, we examined the effect of zVADfmk on the cleavage of Bid observed in response to cytokine plus CHX or following PI3K inhibition. In the presence of CHX or following inhibition of PI3K, TNF, and IL-1 resulted in a clear decrease in the levels of the 22-kDa inactive form of Bid. Following caspase inhibition with zVADfmk, the decrease in p22 Bid was blocked in both cases (data not shown). Similarly, it was also possible to observe dimerization of Bax in response to cytokine following inhibition of PI3K, and this was also blocked by caspase inhibition despite continued cell death (data not shown). These data indicate that treatment with TNF or IL-1 following inhibition of PI3K results in the activation of cathepsin B and that this activation accounts for the loss of mitochondrial function in FIG. 9. Inhibition of cathepsin B blocks cell death in response to TNF or IL-1 following inhibition of PI3K. EC were treated in the presence and absence of CA-074-Me (30 M) with TNF or IL-1 (10 ng/ml) either alone, following pretreatment with LY294002 (50 M), or in combination with CHX (2.5 g/ml). After treatment overnight, detached EC were removed by washing, and substrateattached EC were quantified by Hoescht staining (A). For the analysis of hypodiploid DNA, EC were treated as described in panel A. After treatment overnight (with LY294002) or for 6 h (with CHX) floating and attached EC were harvested, pooled, EtOH-fixed, stained with PI (50 g/ml), and analyzed for hypodiploid DNA by FACS (B). For the analysis of the effect of cathepsin B inhibition on caspase activity, EC were treated as described in panel A. After treatment overnight (LY294002) or for 6 h (CHX) the broad range caspase substrate FAM-VAD-fmk was added for 1 h further before EC were harvested and analyzed for caspase activity by FACS. Data presented are mean fluorescence intensity for FL-1 (C). a manner that is not dependent on the cleavage of Bid and dimerization of Bax. DISCUSSION As previously reported, we observed that EC could be sensitized to the pro-apoptotic actions of TNF and IL-1 by inhibition of protein synthesis with CHX and demonstrated that the cell death observed showed features typical of apoptosis such as DNA degradation, nuclear condensation, and fragmentation and loss of ⌬⌿. Several lines of evidence support the conclusion that the cell death resulting from treatment of EC with TNF or IL-1 in the presence of CHX can be primarily attributed to the mitochondrial-dependent Type I pathway described by Scaffidi et al. (16). The activation of caspases was rapid (detected following only 1 h of treatment) and in agreement with other studies associated with depletion of the endogenous caspase-8 inhibitor c-FLIP (19,22). Furthermore, despite depletion of the pro-form of Bid and loss of ⌬⌿, the activation of the mitochondrial pathway was not essential for the progression of apoptosis in response to cytokine plus CHX as shown by the fact that overexpression of Bcl-2 had no effect or actually decreased cell survival. Bcl-2 expression also had no effect on the activation of caspase-8 indicating that this activation occurred upstream of the loss of mitochondrial function. As expected, blocking caspase activation with zVADfmk prevented all morphological and biochemical parameters of apoptosis. zVADfmk-protected cells remained viable as indicated by effective replating, perhaps the most stringent assay of cell survival.
EC could also be sensitized to the pro-apoptotic actions of TNF and IL-1 following inhibition of PI3K with LY294002. This death pathway occurred with features typical of apoptosis such as DNA degradation, and nuclear condensation and fragmentation. However, caspase activation was delayed and less complete, and death could be blocked by overexpression of Bcl-2. These data initially suggested a Type II amplification pathway of caspase-3 activation. However, nuclear condensation, loss of ⌬⌿, and cell death were still observed during complete caspase inhibition by zVADfmk as detected in an activity assay. Moreover, although it did not prevent death, zVADfmk still caused inhibition of nuclear fragmentation and of DNA degradation. These data suggest that caspases were activated and contributed to some of the morphological and biochemical features of death but that caspases were not necessary for death to occur. It is, however, formally possible that some degree of caspase activation occurs in the presence of zVADfmk that is too slight to measure yet still contributes to cell death. Given the effectiveness of inhibition of cytokine plus CHX treated by zVADfmk, this appears highly unlikely.
Caspase-independent cell death has previously been observed in a variety of other cell types. Caspase inhibition has been ineffective in suppressing anti-CD2 or staurosporine-in- duced death of activated T lymphocytes (24), TNF-plus CHXmediated death of NIH3T3 cells (26), or TNF-mediated cell death of WEHI-S fibrosarcoma cells (25). The activation of non-caspase proteases such as cathepsins, calpains, or granzymes is now emerging as an alternate means to induce cell death (39,40). Cathepsins are lysosomal proteases primarily thought to be involved in the degradation of proteins within the lysosomal compartment. However, several studies have shown translocation of cathepsins from the lysosome to the cytosol during cell death indicating that cathepsins may be able to gain access to cytosolic proteins (41,42). The implication of cathepsin B activation during TNF-mediated cell death in mouse hepatocytes and human fibrosarcoma cells in vitro (25,27,41) prompted us to consider the role of cathepsin B in TNF-and IL-1-mediated cell death observed following inhibition of PI3K in EC. We observed that inhibition of cathepsin B with CA-074-Me prevented loss of ⌬⌿, caspase activation, DNA fragmentation, and cell detachment observed in response to cytokine treatment following inhibition of PI3K. This inhibition occurred without significant effect on the same parameters activated in response to cytokine plus CHX. We also observed translocation of cathepsin B to the cytosol following inhibition of PI3K prior to the activation of caspases. Together, these results indicate that cathepsin B plays a critical role in a caspase-independent, mitochondrial-dependent death pathway in cytokine-treated EC.
We do not know the mechanism for cathepsin B activation in cells treated with TNF or IL-1 plus LY294002. Cathepsin B was liberated from the lysosomes by the actions of LY294002 alone. This agent produced some degree of cell death, but the presence of TNF or IL-1 markedly enhances this effect. We do not know how either LY294002 or cytokine work in this system. Because EC could not be sensitized to cytokine-mediated cell death following inhibition of Akt with K179M Akt, it is unlikely that Akt is involved in this pathway. Guicciardi et al. (41) reported that activation of cathepsin B by TNF plus actinomycin D was likely to be caspase-dependent, because recombinant caspase-8 could cause the release of cathepsin B from purified lysosomes. Our data do not support this conclusion. We observed that loss of ⌬⌿ was not blocked by inhibition of caspases, that transduction of EC with Bcl-2 prevented cytokine-dependent caspase-8 activation, and that the treatment of EC with LY294002 either alone or in combination with cytokine treatment failed to decrease levels of c-FLIP. We therefore conclude that the TNF-or IL-1-mediated death pathway activated in EC following inhibition of PI3K results from cathepsin B activation in a caspaseindependent manner. A recent report has shown that selective lysosomal permeabilization by TNF in hepatocytes is paralleled by sphingosine-mediated lysosomal permeabilization (42). Inhibition of PI3K may sensitize EC to generate sphingosine either by the activation of sphingomyelinase (43) or by de novo biosynthesis (44).
How cathepsin B may result in the activation of a mitochondrial death pathway generating a phenotype with partial nuclear condensation is also undefined. Guicciardi et al. (41) allude to the possibility that cathepsin B activation might result in activation of a mitochondrial death pathway by the cleavage and activation of a cytosolic protein such as Bid. Stoka et al. (45) have also demonstrated cleavage of recombinant mouse Bid following treatment with isolated lysozymes. In contrast to these reports, we observed that caspase inhibition was able to prevent Bid depletion that occurred in response to cytokine treatment either following inhibition of PI3K or protein synthesis. Caspase inhibition was also completely effective in preventing the dimerization of Bax. These observations indicate that Bid cleavage and Bax dimerization are most likely consequences of caspase activation downstream of the mitochondria and not required to initiate cell death. We also failed to detect any increase in p53 as a result of treatment with cytokine plus LY294002. 2 Identification of cathepsin substrates that are pertinent to the caspase-independent death pathway will be an important direction for future studies. FIG. 11. Inhibition of PI3K results in translocation and activation of cathepsin B. For the analysis of the translocation of cathepsin B, cytosolic extracts were prepared by digitonin extraction under conditions that cause selective permeabilization of the plasma membrane without perturbing lysosomes. EC were treated for 3 h with vehicle with or without TNF following pretreatment with vehicle with or without LY294002 (50 M). Cytosolic extracts were analyzed for cathepsin B by Western blotting (A). Similar results were obtained in two other experiments. For the analysis of cathepsin B activity, EC were treated as above either with or without CA-074-Me (30 M) (B). Data are means Ϯ S.E. of three observations. Similar results were observed in one other experiment.