Endothelial cell death induced by tumor necrosis factor-alpha is inhibited by the Bcl-2 family member, A1.

Endothelial cells play a central role in the inflammatory process. Tumor necrosis factor-α (TNF) is a multifunctional cytokine which elicits many of the inflammatory responses of endothelial cells. While TNF directly causes apoptosis of tumor cells and virally infected cells, normal cells are generally resistant. However, most resistant cells, including human endothelial cells, can be rendered susceptible to TNF by inhibiting RNA or protein synthesis. This finding suggests that TNF provides a cell survival signal in addition to a death signal. We have previously cloned a human Bcl-2 homologue, A1, and shown that it is specifically induced by proinflammatory cytokines but not by endothelial growth factors. In this study, we show that retroviral-mediated transfer of the A1 cDNA to a human microvascular endothelial cell line provides protection against cell death initiated by TNF in the presence of actinomycin D. The induction of A1 by TNF in this system is mediated via a protein kinase C pathway. Since TNF signaling has also been shown to proceed via ceramides, we tested whether exogenous ceramides could induce A1. Our findings indicate that ceramides do not induce A1 but do up-regulate c-jun and induce endothelial death. Ceramide-activated endothelial death is also inhibited by A1, suggesting that TNF may initiate divergent survival and death pathways via separate lipid second messengers.


Endothelial cells play a central role in the inflammatory process. Tumor necrosis factor-␣ (TNF) is a multifunctional cytokine which elicits many of the inflammatory responses of endothelial cells. While TNF directly causes apoptosis of tumor cells and virally infected cells, normal cells are generally resistant. However, most resistant cells, including human endothelial cells, can be rendered susceptible to TNF by inhibiting RNA or protein synthesis. This finding suggests that TNF provides a cell survival signal in addition to a death signal. We have previously cloned a human Bcl-2 homologue, A1, and shown that it is specifically induced by proinflammatory cytokines but not by endothelial growth factors.
In this study, we show that retroviral-mediated transfer of the A1 cDNA to a human microvascular endothelial cell line provides protection against cell death initiated by TNF in the presence of actinomycin D. The induction of A1 by TNF in this system is mediated via a protein kinase C pathway. Since TNF signaling has also been shown to proceed via ceramides, we tested whether exogenous ceramides could induce A1. Our findings indicate that ceramides do not induce A1 but do up-regulate c-jun and induce endothelial death. Ceramide-activated endothelial death is also inhibited by A1, suggesting that TNF may initiate divergent survival and death pathways via separate lipid second messengers. TNF 1 is an inflammatory cytokine originally defined by its tumoricidal activity (1,2). Subsequently, it has been shown to evoke multiple biological responses affecting virtually every cell type (3). Considerable attention has recently been paid to the apoptotic pathway elicited by TNF. A series of elegant experiments have defined a death pathway emanating from the TNF receptor 1. Engagement of the TNF receptor results in cell death by recruitment of a complex of proteins to the cell mem-brane including TRADD, FADD/Mort1, and MACH/FLICE which culminates in the activation of cysteine proteases (4 -10). TNF stimulation also results in signal transduction pathways mediated by two lipid second messengers, diacylglycerol and ceramide (3). While ceramide is a transducer of TNF-induced apoptosis (11,12), diglyceride or phorbol ester pretreatment can protect against ceramide-induced cell death (11,13). Ceramides appear to act downstream of FADD/Mort1 since a FADD dominant-negative can block TNF-induced apoptosis but not death mediated by ceramide (7).
Although tumor cells and virally infected cells are susceptible to TNF-induced death, many normal cells are not (1,2). In this respect, human endothelial cells which play a pivotal role in modulating the inflammatory response are not directly killed by TNF (14). However, it has previously been shown that in the presence of RNA or protein synthesis inhibitors, endothelial cells as well as other cells can be rendered sensitive to TNF (14 -16). Conversely, it has also been demonstrated that sensitive cells can be made resistant to TNF challenge by prior sublethal exposure to TNF (15,16). These findings have led to the hypothesis and demonstration of TNF-inducible genes which confer a protective effect on cells. Genes in this category include manganous superoxide dismutase (17), plasminogen activator inhibitor type 2 (18,19), and the zinc-finger protein, A20 (20). Conflicting data have been reported on the effect of the anti-apoptotic protein, Bcl-2 (21). Even in the same cell type, some investigators have reported that Bcl-2 inhibits TNFmediated death while others have claimed that it does not (22,23).
We have recently cloned the human homologue of the Bcl-2 family member, A1, from phorbol ester-stimulated endothelial cells and shown that it is induced by the inflammatory cytokines, TNF and interleukin (IL)-1 (24). This induction is specific in that neither the endothelial growth factors, basic fibroblast growth factor and vascular endothelial growth factor, nor interferon-␥ or transforming growth factor-␤ induce A1. To determine whether A1 belongs to the family of cytoprotective TNF-inducible genes, we generated retroviral constructs of a FLAG-tagged A1 and of Bcl-XL (25) and transduced a microvascular endothelial cell line, HMEC-1 (26), with either of these constructs or the empty vector, LNCX (27). In the presence of actinomycin D, these cells underwent cell death which was inhibited by both A1 and Bcl-XL. TNF induced A1 via a PKC-mediated pathway, while ceramide did not induce A1 but did up-regulate c-jun mRNA. A1 and Bcl-XL also protected HMEC-1 cells from ceramide-mediated apoptosis. Hence, A1 is a Bcl-2 homologue which is induced by TNF and protects against TNF-mediated cell death.

MATERIALS AND METHODS
Reagents-Phorbol 12-myristate 13-acetate (PMA), 4-␤-phorbol, and mezerein were purchased from Sigma. C 2 -ceramide and dihydroceramide were obtained from Biomol and reconstituted in ethanol. TNF was purchased from R&D Systems. Calphostin C, H7, and HA1004 (Calbiochem) were reconstituted in dimethyl sulfoxide. FLAG M2 monoclonal antibody was obtained from IBI Scientific and the Bcl-X rabbit polyclonal antibody from Transduction Laboratories. The horseradish peroxidase-conjugated secondary antibodies used were purchased from Bio-Rad Laboratories.
Gene Transfer-The coding region of Bcl-XL cDNA (25) (provided by L. Boise and C. B. Thompson, University of Chicago) was ligated into the HindIII/HpaI sites of the replication-deficient retroviral vector pLNCX (27) (provided by A. D. Miller). A FLAG octapeptide N-terminaltagged human A1 was generated by PCR using the following primers: sense, 5Ј-CCAGCTAAGCTTCCACCATGGACTACAAGGACGACGAT-GACAAGCAGACTGTGAATTTGGAT-3Ј, antisense, 5Ј-GGTAAAGAA-TTCTCTGGTCAACAGTATTGCT-3Ј, and ligated into the HindIII/HpaI site of pLNCX. The PCR product was sequenced on both strands to confirm the authenticity of FLAG-A1. The viral long terminal repeat drives expression of Neo R while the cytomegalovirus promoter drives transgene expression in pLNCX. Generation of packaging cell lines was performed as described (29). The pLNCBcl-XL, pLNCFLAG-A1 constructs, or pLNCX were transiently transfected into the ecotropic packaging line, PE501, by calcium-phosphate precipitation. Viral supernatants were harvested and used to transduce the amphotropic line PA317 in the presence of 4 g/ml Polybrene. Polyclonal retrovirus-producing cell lines were obtained by selection in 1 mg/ml G418 (Life Technologies, Inc.). Retroviral supernatants from the PA317 cell lines were used to transduce HMEC-1 cells. Following selection in 200 g/ml G418 and expansion, HMEC-1 cells were used in survival studies. Polyclonal HMEC-1 lines were used in order to avoid artifacts due to retroviral integration and also because this cell line does not grow in colonies.
Western Blotting-Total cellular extracts from the transduced cells were prepared by lysing cells in 1 M Tris, 1 M NaCl, 1% Triton X-100, 1 mM 4-(2-aminoethyl)benzenesulfonyl fluoride, 10 g/ml leupeptin, and 10 g/ml aprotinin. Protein from 1 ϫ 10 6 cells was fractionated on 10% SDS-polyacrylamide gel electrophoresis gels and electrotransferred onto nitrocellulose membranes over 1 h at 4°C. Filters were blocked overnight with TBS containing 5% skim milk. Immunostaining steps were performed in TBS with 0.05% Tween 20 and 3% bovine serum albumin at room temperature. Filters were incubated with primary and secondary antibodies for 1 h each. Filters were washed in TBS and 0.05% Tween 20 four times for 10 min between each step and were developed with the ECL reagent (Amersham). Blots were stripped as recommended (Amersham) and reprobed with another antibody.
Viability Assay-For viability assays, transduced or wild type HMEC-1 cells were seeded on gelatin-coated 96-well plates at a density of 15,000 cells/well. By the following day, cells had reached confluence and were washed twice in serum-free medium followed by incubation in TNF (concentrations as indicated) and actinomycin D (1 g/ml) or in C 2 -ceramide or C 2 -dihydroceramide (control for the specificity of ceramide-mediated effects). At various time points, viable cell numbers were estimated by an MTT assay (30). Briefly, medium was removed and replaced with medium containing 1 mg/ml MTT (Sigma) and incubated for 5 h. The medium was then aspirated, and the formazan product was solubilized with dimethyl sulfoxide. Absorbance at 630 nM was subtracted (to reduce background absorbance) from absorbance at 570 nM for each well.
Northern Analysis-Endothelial cells were stimulated for 3 h with the various factors as indicated. Total cellular RNA (15 g) was separated on agarose-formaldehyde gels, blotted onto nitrocellulose filters, and hybridized overnight with random-primed 32 P-labeled probes as indicated. The A1 probe was generated by reverse transcriptase-PCR as described previously (24). The c-jun probe was an approximately 700base pair HindIII/SacII fragment of the murine cDNA (gift of D. Morris, University of Washington, Seattle, WA). The final washing conditions were 0.1 ϫ SSC, 0.1% SDS for 15 min at room temperature. Blots were stripped in boiling water prior to reprobing. A ␤-actin probe (Clontech) was used to confirm equivalent loading of RNA samples.

RESULTS AND DISCUSSION
By virtue of their location between blood and tissue, endothelial cells play a central role in the inflammatory process (31,32). TNF has multiple effects on endothelial cells ranging from structural reorganization to the up-regulation of adhesion molecules and the elaboration of chemotactic factors (31,32). Thus, the integrity of the endothelium during inflammation is of great importance. It has previously been shown that human endothelial cells are not susceptible to TNF toxicity unless coincubated with either cycloheximide or actinomycin D (14). To determine whether human A1, which is inducible by TNF, can abrogate TNF-mediated death, we generated polyclonal FLAG-A1 overexpressing HMEC-1 cell lines. We chose retroviral transduction to achieve overexpression because human endothelial cells are extremely difficult to transfect using standard methods. As a control, we chose to overexpress Bcl-XL (25), another member of the Bcl-2 family, since it is constitutively expressed in cultured endothelial cells (Fig. 1B) 2 and may be one of the constitutive TNF cytoprotectants. Also, Bcl-XL has been shown to protect the MCF7 breast carcinoma cell line from TNF-induced apoptosis (33). Attempts were made to generate clonal cell lines as well, but HMEC-1 cells do not proliferate in colonies, so all experiments on transduced cells were done with polyclonal cell populations.
When a FLAG-A1 construct was overexpressed in HMEC-1 cells, these cells were protected from TNF cytotoxicity. As shown in Fig. 1A, 18-h incubation of HMEC-1 cells in the presence of TNF at various concentrations and actinomycin D resulted in dose-dependent cell death. The fact that about half of the vector-only transduced cells (HMEC-Neo) were still viable suggests the presence of constitutively expressed endogenous inhibitors of TNF cytotoxicity. This level of endogenous protection was also noted in the wild-type HMEC-1 cells (data not shown). One of these cytoprotective molecules may be Bcl-XL since it is constitutively expressed in endothelial cells

FIG. 1. A1 inhibits TNF-mediated endothelial cell death.
Stable polyclonal cell lines of HMEC-1 cells were generated by transduction with LNCFLAG-A1, LNCBcl-XL, or LNCX. A, cells were exposed to TNF and actinomycin D (1 g/ml) for 18 h, and viability was assessed by an MTT assay. Viability is expressed as a proportion of cells treated with actinomycin D only. Results are the mean Ϯ S.E. of an experiment done in triplicate which is representative of four separate experiments. B, Western blots of the cell lines shown above were probed with the FLAG M2 monoclonal antibody or a Bcl-XL polyclonal antibody. C, HMEC-FLAG-A1 or HMEC-Bcl-XL cells were exposed to 50 g/ml cycloheximide, and levels of the overexpressed protein were monitored at various times by Western blot. (Fig. 1B), 2 but not induced by TNF (data not shown). Indeed, overexpression of Bcl-XL resulted in virtually complete protection against TNF (Fig. 1A). The partial protection conferred by A1 may, in part, be due to the shorter half-life of A1 protein.
When HMEC-FLAG-A1 or HMEC-Bcl-XL cells were exposed to cycloheximide (50 g/ml) to inhibit new protein synthesis, for up to 12 h, FLAG-A1 protein was not detectable after 3 to 6 h whereas Bcl-XL levels remained virtually unchanged (Fig. 1C). Thus, after the first few hours of treatment with TNF and actinomycin D, following degradation of existing RNA and protein, there are probably insufficient levels of A1 to provide protection. Although others have shown that murine A1 inhibits apoptosis following IL-3 withdrawal in an IL-3-dependent myeloid cell line (34), human A1 is not induced by endothelial growth factors but rather by proinflammatory cytokines (24). Thus, this model provides a more plausible scenario for the putative function of A1 in endothelial cells.
Because PMA also induces A1 (24), we wondered whether PKC was involved in transducing the TNF signal to up-regu-late A1 expression. Several TNF responses have previously been shown to be mediated via PKC (3). Fig. 2 shows that PMA and TNF induced A1 independently. The highly specific PKC inhibitor, calphostin C (0.1 M), completely blocked TNF-mediated A1 induction. Additionally, H7, another PKC inhibitor, also partially blocked A1 induction, but HA1004, which at the concentrations used (30 M) blocks PKA and PKG but not PKC (35), had no effect. These findings suggest that PKC is involved in the signal to up-regulate A1 expression.
TNF can also signal through another lipid second messenger, ceramide, generated by the hydrolysis of sphingolipids (3). Ceramides also have pleiotropic effects on cells, dependent on cell type and the context in which ceramide is generated. In fact, ceramides have been reported to induce both apoptosis and mitogenesis (11,36,37). Besides TNF, IL-1, ionizing radiation, and chemotherapeutic agents have also been reported to generate ceramides (3, 38 -41). Multiple potential targets of ceramide stimulation have been identified including activation of the atypical PKC (36,42). Hence, we investigated whether exogenous membrane-permeable ceramides could result in A1 mRNA accumulation. Fig. 3 shows that PMA and the nonphorbol PKC activator, mezerein, both induced A1, but the inactive phorbol analogue, 4-␤-phorbol, did not. Again this suggests that PKC is involved in A1 induction. The addition of either C 2 -ceramide or the inactive analogue, C 2 -dihydroceramide, did not result in A1 accumulation when tested over a wide range of concentrations. Recent studies have shown that ceramide-induced apoptosis is associated with induction of c-jun and activation of c-jun N-terminal kinase in human myeloid cell lines and bovine endothelial cells (41,43,44). To determine if c-jun is also up-regulated by ceramides in human endothelial

FIG. 4. A1 inhibits ceramide-mediated endothelial cell death.
A, wild type HMEC-1 cells were exposed to C 2 -ceramide or C 2 -dihydroceramide for 24 h followed by an MTT assay to assess viability. HMEC-1 transductants were exposed to C 2 -ceramide (B) at various concentrations for 24 h or to C 2 -ceramide (C) ( cells and also to verify the membrane permeability of C 2 -ceramide in this system, the Northern blots were probed with a c-jun partial cDNA. A dose-dependent accumulation of c-jun mRNA, which has been shown to be due to increased transcription (44), was noted when HMEC-1 cells were stimulated with C 2 -ceramide, but only slight accumulation was noted at a very high dose of C 2 -dihydroceramide (Fig. 3).
Haimovitz-Friedman et al. (38) have shown that ionizing radiation acts on bovine aortic endothelial cell membranes to generate ceramides and initiate apoptosis. Others have shown that TNF and IL-1 can generate ceramides in human endothelial cells (39,40). The accumulation of c-jun mRNA in our experiments (Fig. 3) suggested to us that ceramides may also initiate apoptosis rather than mitogenesis in human microvascular endothelial cells. Wild type HMEC-1 cells were exposed to C 2 -ceramide or C 2 -dihydroceramide for 24 h, and viability was assessed by MTT assays. As described for bovine endothelial cells, ceramide induced human microvascular endothelial death (Fig. 4A). The toxicity seen with C 2 -dihydroceramide at 50 M has also been noted in other studies with prolonged exposure at higher doses (7,45) and may be due to intracellular conversion of dihydroceramide to ceramide (46) or a nonspecific effect. Bcl-2 and Bcl-XL have been shown to block ceramideinduced death in lymphoid cell lines (47)(48)(49). To determine whether A1 also protects against ceramide-induced death, the HMEC-1 transductants were exposed to C 2 -ceramide at various concentrations. Both A1 and Bcl-XL protected against ceramide-mediated cell death (Fig. 4, B and C). In this instance, the protective effects of A1 and Bcl-XL were similar although the effect of Bcl-XL was slightly more prolonged. Since actinomycin D is not required for ceramide-initiated death, levels of FLAG-A1 would be maintained through the experiment. This continued expression of A1 would help to explain the apparent different efficacies of A1 in TNF/actinomycin D-mediated death compared to death initiated by ceramide alone.
The findings that A1 is induced via a PKC pathway and can protect against ceramide-initiated death are in keeping with those showing that ceramide-induced apoptosis can be attenuated by diglyceride, phorbol esters, or mezerein (11,13). It has also been shown that phorbol esters inhibit endothelial death due to serum and growth factor deprivation and ionizing radiation suggesting that PKC-signaled events are crucial for endothelial survival under various conditions (50,51).
Although TNF-mediated death has gained much attention recently, our study bolsters the postulate that in many cells TNF also activates a cell survival pathway that protects against its apoptotic effects. While the death pathway in endothelial cells does not require new gene or protein expression, the inducible survival pathway requires some gene expression. Two lipid second messengers have been implicated in TNF signaling, diacylglycerol and ceramide. We speculate that in TNF-activated endothelial cells the ceramide pathway propagates death which is independent of new gene expression. We demonstrate that the TNF death pathway can be inhibited via a PKC pathway, in part by up-regulation of the Bcl-2 family member, A1. Whether in HMEC-1 cells TNF induces an increase in ceramides to levels comparable to that achieved by the exogenously added ceramide used in these experiments requires further study.