A1 functions at the mitochondria to delay endothelial apoptosis in response to tumor necrosis factor.

Tumor necrosis factor (TNF) does not cause endothelial apoptosis unless the expression of cytoprotective genes is blocked. We have previously demonstrated that one of the TNF-inducible cytoprotective genes is the Bcl-2 family member, A1. A1 is induced by the action of the transcription factor, NFkappaB, in response to inflammatory mediators. In this report we demonstrate that, as with other cell types, inhibition of NFkappaB initiates microvascular endothelial apoptosis in response to TNF. A1 is able to inhibit this apoptosis over 24 h. We demonstrate that A1 is localized to and functions at the mitochondria. Whereas A1 is able to inhibit mitochondrial depolarization, loss of cytochrome c, cleavage of caspase 9, BID, and poly(ADP-ribose) polymerase, it does not block caspase 8 or caspase 3 cleavage. In contrast, A1 is not able to prevent endothelial apoptosis by TNF over 72 h, when NFkappaB signaling is blocked. On the other hand, the caspase inhibitor, benzyloxycarbonyl-VAD-formylmethyl ketone, completely blocks TNF-induced endothelial apoptosis over 72 h. Our findings indicate that A1 is able to maintain temporary survival of endothelial cells in response to TNF by maintaining mitochondrial viability and function. However, a mitochondria-independent caspase pathway eventually results in endothelial death despite mitochondrial protection by A1.

Activation of endothelial cells by cytokines plays a critical role in the initiation and potentiation of the inflammatory process (1). The inflammatory effects of tumor necrosis factor (TNF) 1 on endothelial cells include the modulation of adhesion molecule expression, secretion of chemokines, and the up-reg-ulation of other cytokines such as interleukin-1 (2,3). Studies over the past few years have also shown that TNF can induce parallel pathways of apoptosis and anti-apoptosis in endothelial cells (4 -7). In general TNF stimulation of endothelial cells does not result in apoptosis unless the expression of new genes is inhibited by using either cycloheximide (CHX) or actinomycin D (Act D) (4 -8). This finding indicates that the TNF death pathway does not require new gene expression but rather activation of proteins that are already expressed in the cell. In contrast, the anti-apoptotic proteins that serve to curb activation of the death pathway are synthesized following TNF stimulation (7,8).
The sequence of events leading to activation of the death pathway by TNF has been well elucidated. Following trimerization of TNF receptor 1 by TNF, apoptosis is activated through the release of silencer of death domains and recruitment of a series of death domain-containing proteins, TRADD, RIP, and FADD, resulting in the formation of a death-induced signaling complex (DISC) (9 -14). DISC formation results in activation of the apical death proteases caspase 8 and/or caspase 10, followed by activation of downstream caspases, and culminating in the cleavage of various cytoplasmic and nuclear substrates (15)(16)(17). Recently, it has been demonstrated that caspase 8 also cleaves the pro-apoptotic Bcl-2 family member, BID (18 -20). The C-terminal fragment of BID translocates to the mitochondria to facilitate the release of cytochrome c from the intermembrane space (18 -20). Cytosolic cytochrome c activates Apaf-1, which oligomerizes caspase 9 leading to the activation of downstream caspases (21). Several cytoprotective genes have been shown to be up-regulated in response to TNF, including manganese-superoxide dismutase, A20, PAI-2, XIAP, c-IAP1 and c-IAP2, IEX-1L, and A1 (also called Bfl-1) (7,(22)(23)(24)(25). We have been particularly interested in A1, as it is a member of the Bcl-2 family that is included in the TNF-inducible set of cytoprotective genes (5,26). We have previously shown that A1 inhibits TNF-initiated apoptosis in the presence of Act D in human microvascular endothelial cells (HMEC) (5). However, in human umbilical vein endothelial cells (HUVEC), death induced by TNF and CHX is not blocked by A1 or Bcl-2 (6). A similar discrepancy in the ability of Bcl-2 to inhibit TNF-mediated apoptosis has been reported (27,28). Although extensive work has been done with Bcl-2 and Bcl-XL to understand their mechanism of action, little is known about A1. Given that A1 is a much smaller molecule than Bcl-2 or Bcl-XL and that it has been shown to have functions distinct from Bcl-2, we were interested in understanding how A1 functions to inhibit apoptosis in endothelial cells (26, 29 -31).
The transcription factor NFB is a key mediator of both the inflammatory response and the cytoprotective response. Inhi-bition of NFB activity in many, but not all, cell types results in the sensitization of cells to TNF-initiated apoptosis in the absence of Act D or CHX (32)(33)(34). We have previously demonstrated that induction of A1 in response to TNF or LPS is dependent on NFB activation, a finding that has subsequently been confirmed by others (35)(36)(37). An HMEC cell line that constitutively expresses an inhibitor of NFB, HMEC-FlagIBmt, does not upregulate A1 following TNF stimulation (35). In this report we demonstrate that inhibition of NFB sensitizes endothelial cells to TNF-induced death. Cotransduction of the HMEC-FlagIBmt cell line with A1 protects endothelial cells from death. A1 does not abrogate activation of caspase 8 or caspase 3 but, consistent with its mitochondrial localization, prevents mitochondrial depolarization, inhibits release of cytochrome c, and activation of caspase 9. In contrast to studies using extracts of Bcl-XL-expressing cells, we find that A1 is able to prevent cleavage of BID in this model. Although maintaining mitochondrial viability initially protects endothelial cells from apoptosis, over 72 h A1-overexpressing cells succumb to cell death due to caspase activation and cleavage of downstream substrates.

Cell Lines
HMEC (38) were provided by the Center for Disease Control and Prevention, Atlanta, GA, and were cultured in MCDB medium supplemented with 10% fetal calf serum and 10 g/ml epidermal growth factor. Cells were maintained at 37°C in 5% CO 2 . Generation of HMEC-Flag-A1, HMEC-FlagIBmt, and HMEC-LNCX cells has previously been described (5). HMEC-FlagIBmt/LXSH and HMEC-FlagIBmt/ HA-A1 cells were constructed by retroviral cotransduction using previously described methods (5). Stable HMEC-1 cell lines were obtained by selection in 100 g/ml hygromycin (Calbiochem) and/or 300 g/ml G418 (Life Technologies, Inc.). Polyclonal HMEC lines were used in order to avoid artifacts due to retroviral integration site.

Apoptosis Assays
Oligonucleosomal Banding-Oligonucleosomal banding was demonstrated by harvesting of total cellular DNA with separation on a 2% agarose gel as described previously (39).
Viability Assay-For viability assays, HMEC were seeded on 96-well plates at a density of 15,000 cells/well. On the following day, cells were incubated in TNF (concentrations as indicated) with or without CHX (50 g/ml) or Act D (1 g/ml). At various time points, viable cell numbers were estimated by 3-(4Ј,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay or neutral red uptake as described (5,40). Viability was expressed as a proportion of cells incubated without TNF.
Hypodiploid Fraction-To assess the proportion of cells with subdiploid DNA content, cells were permeabilized in ethanol and stained with propidium iodide as described previously (39).
Loss of ⌬⌿ m -To measure mitochondrial transmembrane potential (⌬⌿ m ), 5 ϫ 10 5 cells were incubated with 3,3Ј-dihexyloxacarbocyanine iodide (DiOC 6 ) (Molecular Probes) and analyzed for fluorescence on a flow cytometer. The mitochondrial uncoupler, carbonyl cyanide m-chlorophenylhydrazone (Sigma) was used as a positive control for the detection of loss of ⌬⌿ m .
Caspase Activity Assay-IETD-p-nitroaniline (pNA) cleavage was performed with a colorimetric assay kit according to the manufacturer's instructions (R & D Systems). Briefly, 200 g of whole cell lysates from HMEC cells exposed to 10 ng/ml TNF for various times was combined with 200 M IETD-pNA in a 96-well plate and incubated at 37°C. The release of the chromophore by active caspases was quantitated at 405 nm and normalized to untreated cell lysates.

Immunoblotting
To separate intracellular membrane and cytosolic fractions, cells were lysed in 0.025% digitonin, and pellet (membrane) and supernatant (cytosolic) fractions were separated. Fractionation of mitochondria into the membrane fraction was confirmed in each experiment by probing with anti-cytochrome c oxidase. Total cellular extracts were prepared from HMEC, fractionated by SDS-polyacrylamide gel electrophoresis, electrotransferred onto nitrocellulose membranes, and detected by immunoblotting as described (35). Equal loading of lanes was confirmed by Ponceau staining the membranes and in some cases by stripping and reprobing blots with an anti-tubulin Ab. Anti-BID Ab was a gift from Xiaodong Wang (University of Texas, Southwestern). Anti-caspase 9 Ab was provided by Xiadong Wang or purchased from PharMingen. Anticaspase 8 Ab was obtained from Upstate Biotechnology, Inc., and anticaspase 3 Ab was purchased from Upstate Biotechnology, Inc., or Transduction Laboratories. Anti-PARP Ab was provided by Guy Poirier (University of Laval, Quebec, Canada) or purchased from Biomol. Anticytochrome c was purchased from PharMingen and anti-cytochrome C oxidase from Molecular Probes.
Immunocytochemistry-HMEC cultures were stained with the immunogold staining technique for intracellular localization of A1. The monolayers were washed with M199 and fixed in 1:1 acetone:ethanol for 7 min at 4°C. After washing in PBS containing 1% bovine serum albumin and 1% normal goat serum (PBS/BSA/NGS, Sigma), cultures were incubated with the primary antibody (anti-FLAG/M5) for 60 min at a final concentration of 5 g/ml in PBS containing 5% BSA and 4% NGS. After washing, the monolayers were incubated with the secondary antibody (Auroprobe TM LM goat anti-mouse IgG conjugated to 5-nm gold particles) at 1:40 dilution for 90 min, washed repeatedly with buffer and double-distilled H 2 O, incubated in silver enhancing solution, and counterstained with Giemsa. Controls included HMEC-LNCX cells and cultures incubated with control buffer or normal mouse IgG instead of primary Ab. Immunoelectron Microscopy-Monolayers were washed with 0.1 M phosphate buffer and fixed in 2% paraformaldehyde and 2.5% glutaraldehyde in phosphate buffer for 1 h at 4°C. Following washing with 0.1 M phosphate buffer, cells were post-fixed in 1% OsO 4 in 0.1 M phosphate buffer for 40 min at 4°C, washed, block-stained in uranyl magnesium acetate, dehydrated, and embedded in Epox 812. Ultrathin plastic sections mounted on nickel grids were incubated with saturated solution of sodium-m-periodate (NaIO 4 ) for 30 min, washed with double-distilled H 2 O, incubated with 0.1 N HCl for 10 min, and then with Tris/Tween 20 buffer containing 2% BSA and 5% NGS. The cells were then incubated with the primary antibody (anti-FLAG/M5) at 1:100 dilution in Tris/ BSA containing 1% NGS for 60 min and then with the secondary antibody (Auroprobe TM EM goat anti-mouse IgG conjugated to 10-nm gold particles) at 1:20 dilution for 60 min. Following washing, sections were counterstained with lead citrate and viewed under a Zeiss EM10.

A1 Inhibits TNF-initiated Death When NFB Activation Is
Blocked in Endothelial Cells-We have previously shown that A1 inhibits TNF-induced apoptosis in HMEC when Act D is used to block RNA transcription (5). However, A1 does not block apoptosis in HUVEC when CHX is used to inhibit protein synthesis (6). The reason for this may be the use of different types of endothelial cells. Alternatively, the apparent discrepancy in A1 function may be due to the use of Act D to block RNA transcription in one case and in the use of CHX to inhibit protein translation in the other. Because of these discrepant results regarding the protective effect of A1 in endothelial cells, we compared death induced by TNF and CHX versus that induced by TNF and Act D in the same type of endothelial cell. In HMEC we found that FLAG-tagged A1 was not able to inhibit TNF-induced death in the presence of CHX (Fig. 1A) but did protect against TNF/Act D-induced death (Fig. 1B). Thus, in order to avoid using CHX or Act D, we employed an HMEC cell line that constitutively expresses a mutant IB␣ (IBmt) that acts as a super repressor of NFB to study TNF-induced endothelial apoptosis (35). We have shown that TNF induction of several anti-apoptotic proteins, including A1, A20, and manganese-superoxide dismutase (35), 2 is blocked in the HMEC-2 X. Hu and A. Karsan, unpublished results.

FIG. 2. IBmt sensitizes HMEC cells to TNF-induced apoptosis.
A, oligonucleosomal banding. Total cellular DNA was isolated from HMEC cells transduced with LNCIBmt or the empty vector LNCX after incubation with (ϩ) or without (Ϫ) TNF (10 ng/ml, 18 h) and separated on a 2% agarose gel. Molecular weight (MW) corresponds to 100-base pair ladder. B, HMEC cells stably transduced with LNCFlagIBmt (q) or the empty vector (E, LNCX) were exposed to indicated doses of TNF for 16 h. Cell viability was measured by neutral red uptake and expressed as a proportion of untreated cells. Results are the mean Ϯ S.E. of an experiment done in triplicate. Data shown are representative of six independent experiments.

FIG. 3. A1 inhibits TNF-induced death of IBmt-transduced HMEC cells.
A, the HMEC LNCFlagIBmt cell line was stably transduced with HA-A1 (q) or empty vector (E, LXSH), and cells were exposed to various concentrations of TNF for 18 h. Cell viability was measured by neutral red uptake and expressed as a proportion of untreated cells. Results are the mean Ϯ S.E. of an experiment done in triplicate. Data shown are representative of four independent experiments. B, apoptosis was measured independently by treating cell lines with TNF for 18 h and quantitating the hypodiploid peak. Results are representative of three independent experiments. C, expression of FlagIBmt and HA-A1 in HMEC cotransduced with Flag-IBmt/LXSH or Flag-IBmt/HA-A1 by immunoblot.
FlagIBmt cell line. The HMEC-FlagIBmt cell line was sensitized to TNF-mediated apoptosis in comparison to vectortransduced cells (HMEC-LNCX), as demonstrated by DNA fragmentation into an oligonucleosomal pattern ( Fig. 2A), neutral red uptake (Fig. 2B), or MTT viability assay (data not shown). To determine whether A1 was capable of preventing death by TNF without the confounding presence of CHX or Act D, we cotransduced the HMEC-FlagIBmt cell line with HA- tagged A1 or the empty vector (LXSH). We found that cotransduction of HA-A1 was sufficient to block TNF-mediated death in these cells as demonstrated by neutral red uptake (Fig. 3A) or MTT assay (not shown). Protection by A1 against apoptosis was confirmed by enumerating the proportion of cells in the hypodiploid fraction, as assessed by flow cytometric analysis of propidium iodide-stained DNA (Fig. 3B). Expression of FlagIBmt and HA-A1 in these cell lines is shown in Fig. 3C.
A1 Functions at the Level of the Mitochondria-Anti-apoptotic members of the Bcl-2 family contain a C-terminal transmembrane domain that allows localization to intracellular organelles such as the mitochondria (41,42). In contrast to other members of the Bcl-2 family, A1 has three charged residues at the site of this putative transmembrane domain (26). Thus to determine whether A1 localizes to mitochondrial membranes despite the lack of a typical transmembrane domain, we performed immunogold staining visualized by light microscopy (Fig. 4A) or electron microscopy (Fig. 4B) on HMEC-FlagA1 cells using an anti-FLAG Ab. As seen in Fig. 4, gold particles demonstrate a particulate distribution that mainly localizes to the mitochondria in these cells. Occasional particles are also seen in the cytosol (Fig. 4B). Only a rare randomly distributed particle is deposited in control cells, and no labeling is present when the primary Ab is omitted (data not shown). To confirm membrane localization in the cotransduced HMEC-FlagIBmt/ HA-A1 cells, cytosolic and intracellular membrane fractions were prepared and subjected to immunoblotting with the anti-HA Ab. Fig. 4C confirms that the majority of A1 is present in the membrane fraction with a small amount also present in the cytosol.
Induction of the mitochondrial permeability transition (MPT) has been implicated in the apoptotic process. The MPT represents an abrupt increase of permeability of the inner mitochondrial membrane to solutes. The rapid increase in permeability associated with the MPT causes uncoupling of oxidative phosphorylation and a reduction in ⌬⌿ m . To investigate further the function of A1 at the mitochondria, we tested whether TNF treatment elicits a decrease in ⌬⌿ m in HMEC-FlagIBmt cells using the lipophilic fluorochrome, DiOC 6 , which partitions to mitochondria (43). Loss of ⌬⌿ m is detected as a decrease in fluorescence. As seen in Fig. 5 TNF elicits a loss of ⌬⌿ m when NFB activation is blocked (HMEC-FlagIBmt/LXSH), and A1 is able to prevent this loss of ⌬⌿ m . As a positive control, the mitochondrial uncoupling agent mchlorophenylhydrazone was able to cause a complete collapse of ⌬⌿ m (data not shown). Similar results were seen with a second dye, JC-1, that can monitor ⌬⌿ m (data not shown).
A1 Inhibits Cytochrome c Release and Caspase 9 Cleavage, but Not Caspase 8 or Caspase 3 Cleavage, at Early Times following TNF Stimulation-Recently, it was shown that Bcl-XL was able to inhibit cytochrome c release from the mitochondria but not caspase 8 cleavage (44,45). Given that A1 localizes to the mitochondria and is able to inhibit mitochondrial depolarization, we postulated that it would also block cytochrome c release. Fig. 6A shows that the A1 is able to inhibit cytochrome c release and caspase 9 cleavage, but as described for Bcl-XL, caspase 8 activation was unaltered (Fig.  6B). Caspase 3 cleavage was also only slightly delayed in HMEC-FlagIBmt/HA-A1 cells (Fig. 6B). Despite early activation of caspase 3, cleavage of the downstream substrate, poly-(ADP-ribose) polymerase (PARP), was significantly inhibited by A1 (Fig. 6C), consistent with the maintenance of viability by A1 (Fig. 3A). Surprisingly, cleavage of BID, which transmits the death signal from the DISC to the mitochondria, was also inhibited by A1 (Fig. 6D). Although IETD-pNA is not an entirely specific substrate (46), we estimated the degree of caspase 8 activity, by measuring IETD-pNA cleavage activity of extracts from TNF-stimulated cells. Fig. 7 shows that caspase activity increases to similar degrees up to 4 h following stimulation, following which caspase activity is attenuated in A1expressing cells compared with the vector-transduced control.
A1 Delays, but Does Not Abolish, Endothelial Apoptosis-As seen in Fig. 3A, A1 was able to inhibit apoptosis induced by TNF when NFB activation is inhibited. In another study, where apoptosis was initiated using TNF in the presence of CHX, Bcl-XL was not able to block cell death, although cytochrome c release was abrogated (20). In this and other previous studies, the use of CHX or Act D has necessitated that assessment of cell viability be done at early times. Because our current model does not require the use of CHX or Act D, which cause death themselves over a prolonged period, we were able to assess TNF apoptotic effects and assay for cell viability in response to TNF stimulation at late time points. Since we were able to detect caspase 3 activation even in A1-expressing cells, we wondered whether apoptosis could still proceed with delayed kinetics. This would be possible through direct activation of caspase 3 by caspase 8 and subsequent cleavage of downstream substrates. Indeed, caspase 8 and 3 are cleaved to a similar degree at later times (Fig. 8A). Although PARP cleavage is apparent by 24 h (Fig. 8B), A1 continues to prevent cleavage of BID (Fig. 8C). By 72 h A1-expressing cells release cytochrome c into the cytosol (Fig. 8D), even though A1 protein is still present in the membrane fraction (Fig. 8E). Over 48 -72 h, A1-expressing HMEC-FlagIBmt cells also lose viability following TNF stimulation as measured by neutral red uptake (Fig. 9A). Interestingly, when viability is measured in parallel using an MTT assay, which predominantly measures mitochondrial function, A1-expressing cells appear to maintain viability (Fig. 9B) (47). The findings suggest that A1 maintains mitochondrial function/viability, but continuing caspase activity causes cell death independently of the mitochondria. Caspase activity is required for TNF-induced death regardless of whether a mitochondria-dependent or -independent pathway is utilized, as demonstrated by the results showing that the broad spectrum caspase inhibitor, ZVAD-fmk (50 M), was sufficient to prevent apoptosis of HMEC-FlagIBmt/ LXSH (and HMEC-IBmt/HA-A1) cells over 72 h (Fig. 9C).

DISCUSSION
In contrast to Bcl-2 and Bcl-XL, the function of A1, a TNFinducible member of this family, has not been well studied (41,42). A1 is a smaller protein that is lacking the loop region in Bcl-2 and Bcl-XL, which is thought to regulate the activity of these molecules (26,30). In the BH4 domain, A1 has minimal homology to Bcl-2/Bcl-XL and a series of glutamine residues that are not present in other family members (26,30). As well, because a typical transmembrane domain is not present at the C terminus of A1, the subcellular localization of this molecule has not been entirely clear (26,30). One study shows that A1 has a cytoplasmic distribution (29). Differences in function between A1 and Bcl-2 have been noted. Unlike Bcl-2, A1 cooperates with the E1A antigen to provide a potent transforming capability in vitro (29). Murine A1 permits myeloid differentiation to granulocytes, whereas Bcl-2 inhibits this process (31). Finally, it has been suggested that A1 may not have the antiproliferative capacity that has been reported for Bcl-2 (30).
We have previously demonstrated that A1 is up-regulated by TNF and LPS through the activation of NFB in endothelial cells (35). Constitutive expression of A1 protects against TNFmediated endothelial death (5). Interestingly, although A1 is protective against TNF-initiated death in the presence of Act D, it cannot prevent apoptosis mediated by TNF and CHX ( Fig.  1) (6). Tang and colleagues (48) have recently demonstrated that CHX-induced T cell death involves FADD-directed activation of caspase 8 and caspase 3. Thus, the presence of CHX may potentiate a mitochondria-independent apoptotic pathway through the direct activation of caspase 3 by caspase 8, which is not inhibitable by A1. On the other hand, the use of Act D may potentiate a mitochondrial death pathway, which is inhib-itable by A1. This observation may explain why in another study Bcl-XL was able to inhibit cytochrome c release, but not cell death, since these investigators used TNF and CHX to trigger apoptosis (20). There have been other contradictory reports regarding protection against TNF-induced apoptosis by Bcl-2 family members (27,28). Whereas cell type differences may partially account for these discrepancies, the use of Act D or CHX as a sensitizer is likely another factor in the reported results.
The above findings for TNF-induced apoptosis can be compared with those described for Fas-induced death. Depending on the cell type, Bcl-2 may or may not block Fas-initiated apoptosis (49). Type I cells show strong activation of caspase 8 by Fas (49). In contrast, type II cells only show weak activation of caspase 8, and the activation of caspase 8 and 3 occurs only following mitochondrial depolarization (49). Bcl-2 inhibits Fasinduced death in type II but not type I cells (49). Interestingly, CHX does not affect DISC formation in type II cells although it sensitizes these cells to Fas-mediated apoptosis likely by synergizing with the mitochondrial death pathway (50). Thus, the CHX effect may depend on cell type as well as apoptotic trigger. Unfortunately, we cannot verify the Fas effect in HMEC-1 cells, since these cells do not express Fas (51).
Because we were interested in studying A1 function, we attempted to avoid the confounding effects of CHX or Act D, by using a model where HMEC-1 cells are sensitized to TNFinitiated apoptosis by the inhibition of NFB using a super repressor IB␣S32/36A (IBmt) (Fig. 2) (52). IBmt cannot be serine-phosphorylated and hence is not degraded, thereby retaining NFB in the cytoplasm (52). Cells that overexpress this mutant (HMEC-FlagIBmt) do not up-regulate A1 in response to TNF or LPS (35). HMEC-FlagIBmt cells were cotransduced with either an HA-tagged A1 or the empty vector (LXSH). When NFB activation is blocked, the onset of apoptosis can be inhibited by A1 (Fig. 3). Although the degree of caspase 8 cleavage is similar whether A1 is expressed or not, caspase 3 activation is slightly delayed in A1-expressing cells. Furthermore, A1 clearly inhibits mitochondrial depolarization, cytochrome c release, caspase 9 activation, BID cleavage, and PARP cleavage. Taken with the mitochondrial localization of A1, the above findings suggest that A1 can prevent mitochondrial events of apoptosis but not the direct activation of caspase 3 by caspase 8, which presumably is a cytosolic event (Fig. 10). Others (53) have also suggested that A1 may function at the mitochondria, because it is able to inhibit apoptosis caused by etoposide, which signals a mitochondrial death pathway.
Given that caspase 3, which is responsible for cleavage of downstream cellular substrates as well as activation of several other caspases, is still activated in the presence of A1, it is difficult to imagine that A1-expressing cells are protected from apoptosis indefinitely (54,55). In previous studies, TNF-induced death could only be studied in the time frame of 24 h due to the concomitant addition of Act D or CHX, which are toxic in their own right. By using the inhibition of NFB to sensitize cells to TNF-initiated apoptosis, we were able to confirm that A1 only delays but does not abolish the apoptotic program when examined over 72 h. Thus by maintaining the viability of the mitochondria, A1 appears to slow the auto-amplification loop of the caspase cascade. Interestingly, the activation of caspases is eventually sufficient to overcome the mitochondrial protective effect of A1, as demonstrated by the release of cytochrome c at 72 h. Also, A1-expressing cells start to lose viability at 48 h, whereas cytochrome c release is not seen until 72 h, suggesting that the dismantling of cellular caspase substrates is sufficient to cause death even while the mitochondria are still functional.
BID has been shown to be cleaved in vitro by caspases 1, 2, 3, 7, and 8 and granzyme B (18 -20, 56). BID appears to be a preferred substrate for caspase 8, and low level caspase 8 activation is sufficient to cleave BID while leaving caspase 3 intact (56). However, despite the activation of caspase 8 and caspase 3, A1 is able to prevent TNF-induced BID activation in this model. This finding is in contrast to that described for Bcl-XL in Fas-induced death of MCF7 cells (18). Since BID cleavage and translocation of the C-terminal fragment is felt to transduce the TNF and Fas signal to mitochondria, it is surprising that A1 is able to prevent BID cleavage (18 -20, 56). It is possible that the delay of activation of the caspase cascade prevents complete processing of BID. Whereas caspase 8 and 3 may activate small amounts of BID, our findings suggest that mitochondrial events, such as the activation of caspase 9, are probably required for significant processing of BID and amplification of the death signal in this endothelial cell system. A1 is only transiently induced by TNF in endothelial cells (26). We speculate that A1 expression may be induced to provide temporary mitochondrial protection as the cellular level of caspase activation increases during inflammation (57,58). Protection at the level of the mitochondria would permit caspase activation for non-apoptotic functions but still prevent cell death transiently in the context of inflammation. It is important to note that other cytoprotective proteins are also induced by TNF, and these molecules may function to block directly caspase activation (7,(22)(23)(24)(25). Nevertheless, caspase 1 is activated by TNF and bacterial lipopolysaccharide and is required for processing pro-interleukin-1 to its mature form (57,59). Although caspase 1 is felt to be required for inflammatory cytokine processing rather than apoptosis, it is able to cleave BID, and thus presumably promote mitochondrial depolarization (56, 60 -62). Thus, one could speculate that added mitochondrial protection, by transient A1 up-regulation, permits physiologically important caspase 1 activation during inflammation.
The broad spectrum caspase inhibitor, ZVAD-fmk, can completely abolish TNF-induced cell death over 72 h (Fig. 9). Hence, regardless of whether TNF-induced death proceeds mainly by a mitochondrial route or a mitochondria-independent pathway, caspases are required. To return to our previous speculation, despite the induction of multiple anti-apoptotic proteins, some caspases are still activated following stimulation by inflammatory mediators (63). Although these caspases are felt to have only a minimal impact in apoptotic pathways, their activation is sufficient to cause cell death (60,61). Therefore, we propose that the expression of cytoprotective proteins, such as A1, that function at the mitochondria are important for providing additional survival capability in inflammatory situations. However, in the context of excessive or prolonged inflammation, this protective role is insufficient to prevent apoptosis and may result in the cell death reported in these inflammatory processes (58).