Tumor Necrosis Factor α Induces BID Cleavage and Bypasses Antiapoptotic Signals in Prostate Cancer LNCaP Cells

INTRODUCTION

The most effective treatment available for metastatic prostate cancer is androgen ablation (1) . The majority of prostate epithelial cells undergo apoptosis after being deprived of androgen. However, prostate cancer cells can often overcome the requirement for androgen and become resistant to other forms of therapy (2 , 3) . Apoptosis is the predominant mechanism by which cancer cells die when subjected to chemotherapy or irradiation (4 , 5) . Thus, resistance to these therapies may be due, at least in part, to the development of effective antiapoptotic mechanisms within prostate cancer cells (6 , 7) . One such mechanism is increased expression of antiapoptotic proteins of the Bcl-2 family. Increased levels of Bcl-2, Bcl-X_L, and Mcl-1 are found in prostate tumors of high pathological grade (8) , and expression of Bcl-2 is sufficient to develop androgen independence in experimental models (9) .

Another mechanism allowing escape from apoptosis is activation of survival signal transduction pathways in prostate cancer cells (10) . Recently, several survival signaling pathways have been described. The best characterized pathway is signaling via activation of PI3K 3 and the protein kinase Akt. When activated, Akt can directly phosphorylate and inactivate the proapoptotic molecule Bad or signal through the transcription factors Forkhead and nuclear factor κB (11) . There is ample evidence that other signaling pathways can be used by steroid hormones, growth factors, and neuropeptides to inhibit apoptosis, but their molecular nature is largely unknown (2 , 5 , 12 , 13) .

In LNCaP prostate cancer cells, Akt is constitutively active as a result of a frameshift mutation in the PTEN gene, which encodes a phosphatase that inactivates the lipid products of PI3K. Decreased or complete loss of PTEN expression is observed in a significant portion of advanced prostate cancers (14, 15, 16, 17) . It is not surprising that cells lacking PTEN are less sensitive to anticancer drugs whose mechanism of action is based on induction of apoptosis.

When LNCaP cells are treated with inhibitors of PI3K and deprived of survival factors, they spontaneously undergo apoptosis. However, treatment with EGF or androgen can protect cells from apoptosis, although Akt activity remains inhibited (10 , 18) . This indicates the existence of one or more Akt-independent survival signaling pathways in LNCaP cells.

The resistance of prostate cancer cells to traditional forms of therapy may be explained by multiple mechanisms protecting these cells from apoptosis. Understanding the molecular basis of these survival pathways and the ways they intersect with the apoptotic cascade and identifying agents that can bypass these signals will likely be necessary for the development of effective therapies for prostate cancer.

Two major arms of proapoptotic signaling leading to activation of effector caspases have been described (see the diagram in Fig. 7 ⇓ ). One, a mitochondrial arm, is engaged by stress, some anticancer drugs, and withdrawal of survival factors. This stress-activated arm induces release of cytochrome c (and possibly other proteins) from mitochondria, leading to activation of caspase 9. The other proapoptotic arm is preferentially activated by ligands of death receptors (e.g., TNF-α) and can operate by recruitment of caspase 8 or caspase 10 (19) . In some cell types, induction of apoptosis by ligands of death receptors also requires release of cytochrome c from mitochondria and activation of caspase 9. This is achieved by cleavage of BID with caspase 8. After being cleaved, the COOH-terminal fragment of BID translocates to mitochondria and induces cytochrome c release (20) . After being activated, “initiator” caspases 8, 9, and 10 can cleave and activate “effector” caspases 3, 7, and 6 (21) .

To determine the subset of proapoptotic molecules targeted by EGF- induced survival mechanisms, we assessed the ability of EGF to inhibit apoptosis induced by factors that operate via different arms for proapoptotic signaling. The PI3K inhibitor LY294002 and TNF-α were used to induce apoptosis via the mitochondrial and death receptor arms, respectively. We found that EGF can protect LNCaP cells from apoptosis induced by LY294002 but not from apoptosis induced by TNF-α. TNF-α induced BID cleavage, cytochrome c release, and apoptosis under conditions in which Akt is active and could bypass the antiapoptotic effects of EGF. In LNCaP cells ectopically expressing BID, proapoptotic signals induced by LY294002 and TNF-α converge at the level of cytochrome c release, upstream of caspase 9 activation. The inability of EGF to inhibit TNF-α-induced apoptosis in LNCaP cells expressing BID indicates that EGF-induced survival signals operate on the mitochondrial pathway upstream of cytochrome c release. The capacity of TNF-α to bypass survival signals makes proapoptotic signaling via death receptors a promising avenue for therapeutic intervention in prostate cancer cells.

MATERIALS AND METHODS

Cell Lines and Plasmids.

LNCaP cells obtained from Leland Chung (University of Virginia, Charlottesville, VA) were routinely maintained in RPMI 1640 supplemented with 10% fetal bovine serum. For experiments described in this article, we used cells between passage 35 and 43.

Expression vector encoding proteolytically inactive mutant (C287A) of caspase 9 has been described previously (22) ; expression vector of GFP was purchased from Clontech (Palo Alto, CA).

C-DNAs coding for NH₂-terminally Flag-tagged truncated forms of N-BID (amino acids 1–59) and C-BID (amino acids 60–195) were generated by PCR using mouse BID as a template. To create N-BID, c-DNA was amplified with Pfu polymerase using specific primers introducing BamHI and EcoRI restriction sites (N-BID forward, 5′-ATACGGATCCGACTCTGAGGTACGCAACGGTTCC-3′, N-BID reverse, 5′-TGAATTCAGTCTGTCTGCAGCTCGTCTTCGAGG-3′, restriction sites are underlined). PCR product was digested with BamHI and EcoRI and cloned into the pcDNAFlag2AB vector (23) digested with the same enzymes. Similarly, to create C-BID, c-DNA was amplified with Pfu polymerase using specific primers introducing BglII and EcoRI restriction sites (C-BID forward, 5′-TTGAAGATCTGGAAGCCAGGCCAGCCGCTCCTTC-3′, C-BID reverse, 5′-TAAGGAATTCAGTCCATCTCGTTTCTAACCAAGT-3′, restriction sites are underlined). PCR product was digested with BglII and EcoRI and cloned into the pcDNAFlag2AB vector digested with BamHI and EcoRI. Accuracy of cloning was confirmed by DNA sequencing and by Western blot analysis of expressed proteins.

Transfection.

Cells were plated as described under “Apoptosis.” The next day, growth medium was changed, and 1 day later, cells were transfected as follows: medium was changed to serum-free RPMI 1640 (referred to hereafter as “medium”); in separate vials, 2 μg of DNA were mixed with 300 μl of medium, and 6 μl of LipofectAMINE (Life Technologies, Inc., Gaithersburg, MO) were mixed with 300 μl of medium; 20 min later, solutions of LipofectAMINE and DNA were mixed together. After 40 min, 1.4 ml of medium were added, and 5 min later, medium was replaced by transfection mixture. One ml of transfection mixture was used per well in a 6-well plate, 2 ml/6-cm dish and 5 ml/10-cm dish. Cells were kept in the transfection mixture for 6 h, which then was replaced by medium with 10% fetal bovine serum. For apoptosis experiments, medium was changed to serum-free medium 18 h later, and after 24 h, apoptosis was induced.

Antibodies.

Antibodies to mouse BID were from Stanley Korsmeyer (Harvard University, Cambridge, MA), antibodies to human BID were from BioVision (Palo Alto, CA), anti-Flag M5 was from Sigma (St. Louis, MO), anti-HA was from BABCO (Berkley, CA), antibodies to phospho-Akt and Akt were from New England Biolabs (Beverley, MA), antibody to phospho-ERK was from Calbiochem (San Diego, CA), anti-phosphotyrosine 4G10 was from Upstate Biotechnology, Inc. (Lake Placid, NY), antibody to caspase 7 (active form) was from Junying Yuan (Harvard University, Cambridge, MA), antibody to caspase 8 (proform) was from Santa Cruz Biotechnology, antibody to caspase 3 (CM-1; active form) was from Idun Pharmaceuticals (La Jolla, CA), and antibody to GFP was from Clontech. Secondary fluorophore-conjugated goat antimouse or goat antirabbit antibodies used for indirect immunofluorescence were from Jackson Immunoresearch Laboratories (West Grove, PA), secondary horseradish peroxidase-conjugated antibodies used for Western blot detection were from Amersham Pharmacia Biotech (Piscataway, NJ). For Western blotting and immunofluorescence on fixed cells, primary antibodies were diluted 1:1000 or according to recommendation of the supplier.

Apoptosis.

LNCaP cells were plated at 3 × 10⁵ cells/well on 6-well plates, 6 × 10⁵ cells/6-cm dish, or 1.2 × 10⁶ cells/10-cm dish. In 2 days, medium was changed to serum-free medium for another 24 h. The cells were then treated with 3 μg/ml CX (Sigma), and after 30- 60 min, 100 ng/ml TNF-α (Upstate Biotechnology, Inc.) or 25 μm LY294002 (Sigma) was added. EGF (100 ng/ml; Upstate Biotechnology, Inc.) or caspase inhibitor Z-Val-Ala-Aspfluoromethylketone (40 μm; Z-VAD-fmk; Bachem, King of Prussia, PA) was added immediately.

Routinely, analysis of apoptosis was done by measuring caspase activity in cell lysates with fluorogenic substrate Ac-Asp-Glu-Val-Asp-amino-4-methyl-coumarin from Upstate Biotechnology, Inc. or by counting the percentage of fragmented nuclei. To measure caspase activity, adherent and floating cells were combined and lysed in NPLB (1% NP40, 150 mm NaCl, 20 mm HEPES, 1 mm EDTA, 1 mm DTT, and 5 μg/ml aprotinin, pepstatin, and leupeptin). Subsequent analysis was done according to the instructions provided with Ac-Asp-Glu-Val-Asp-amino-4-methyl-coumarin.

For microscopic analysis, cells were fixed with 10% formalin in PBS (pH 7.4) for 15 min, washed twice with PBS, and stained with 0.2 μg/ml DAPI in PBS-T. The percentage of fragmented nuclei was determined by counting 100 nuclei in randomly chosen fields. Each sample was counted at least twice, blinded. In most cases, results were independently verified by a second operator. Error bars represent the SD between counts of the same sample. All experiments were repeated at least three times.

Apoptosis in a Population of Transiently Transfected Cells.

To visualize the subpopulation of transfected cells, cDNA for GFP was mixed with the DNA of interest in a ratio of 1:10. Coexpression of GFP with the protein of interest was verified by immunofluorescence and by fluorescence-activated cell-sorting analysis of transfected cells. Quantitation of apoptosis was done by counting the percentage of fragmented nuclei in GFP-expressing cells using a fluorescence microscope.

Immunofluorescence.

Cells were plated on glass coverslips in 6-well plates and treated and fixed as described above. The cells were then permeabilized with 0.1% NP40, 1% sodium citrate in water for 2 min at 4°C and blocked in 2.5% goat serum in PBS-T for 30 min at 37°C. Blocking solution was used to prepare a 1:300 mixture of mouse-anti-cytochrome c antibodies from PharMingen (65971A; San Diego, CA) and a 1:3000 mixture of rabbit anti-active caspase 3 antibodies (CM-1); subsequent procedures were done at room temperature. Incubation with primary antibodies was continued for 3 h. The cells were then washed three times with PBS-T and incubated with a mixture of goat antirabbit antibodies conjugated with Texas Red and goat antimouse antibodies conjugated with fluorescein, diluted in blocking solution at a ratio of 1:300 each. One h later, cells were washed with PBS-T, stained with DAPI, and mounted on glass microscope slides. Image analysis was done on a Nikon microscope equipped with a digital camera and software from Inovision (Durham, NC).

Other experimental procedures were as described previously (10 , 24 , 25) .

RESULTS

Apoptosis Induced by TNF-α Cannot Be Inhibited by Akt or EGF.

We and others have demonstrated previously (18 , 26) that Akt is constitutively active in LNCaP cells, and this is necessary and sufficient to protect cells from apoptosis when deprived of survival factors: LNCaP cells do not lose viability when placed in serum-free conditions but rapidly undergo apoptosis when kept under serum-free conditions and treated with the PI3K inhibitor LY294002. We also reported that EGF can rescue LNCaP cells from LY294002-induced apoptosis via a mechanism that is independent of PI3K and Akt activity (10) . To assess the breadth of the spectrum of proapoptotic stimuli that EGF and Akt can protect cells from, we looked at the effect of TNF-α on apoptosis in LNCaP cells. Apoptosis was assessed by counting the percentage of fragmented nuclei (Fig. 1, A and B) ⇓ , measuring caspase activity (Fig. 1C) ⇓ , and detecting the cleavage of PARP protein (Fig. 1D) ⇓ . As evident in Fig. 1 ⇓ , TNF-α was able to induce apoptosis in LNCaP cells even when Akt was not inhibited with LY924002. In addition, EGF failed to protect cells from TNF-α-induced apoptosis.

TNF-α Does Not Inhibit Signaling Downstream of EGFR.

It was possible that TNF-α escaped survival effects of EGF by inhibiting signaling from the EGFR. To address this possibility, we compared activation of signaling pathways in cells treated with TNF-α or vehicle followed by EGF stimulation. TNF-α did not produce detectable changes in EGF-induced activation of ERK, p70S6 kinase, and overall levels of phosphorylation on tyrosine (Fig. 2) ⇓ . Also, TNF-α did not inhibit constitutively active Akt. Thus, it is more likely that TNF-α bypasses survival signals from EGF and Akt rather than neutralizing them.

TNF-α Induces BID Cleavage and Cytochrome c Release from Mitochondria.

To elucidate the mechanistic basis for the inability of Akt or EGF to protect cells from TNF-α-induced apoptosis, we examined the pathway by which TNF-α induces apoptosis in LNCaP cells. Two mechanisms for induction of apoptosis by TNF-α and other death receptor agonists have been described. Cells are designated as type I or type II (27) , depending on the mechanism that is preferentially used. In type I cells, activation of death receptors leads to direct cleavage and activation of effector caspases by caspase 8. In type II cells, caspase 8 cleaves BID, which then triggers release of cytochrome c and possibly other factors from mitochondria, leading to activation of caspase 9 and effector caspases.

We used antibodies recognizing the cleaved form of PARP, the proform of caspase 8, or full-length BID to see whether these proteins are cleaved in response to proapoptotic agents. As shown in Fig. 1D ⇓ and Fig. 3A ⇓ , treatment with either TNF-α or LY294002 can activate effector caspases 3 and 7 and induce PARP cleavage. However, disappearance of full-length BID and procaspase 8 was evident only in TNF-α-treated cells. These result are consistent with previous reports that show activation of caspase 8 by TNF-α and activation of caspases 3 and 7 in LNCaP cells treated with TNF-α or lovostatin (28, 29, 30) . Thus, they indicate that proapoptotic signals induced by LY294002 and TNF-α activate a similar repertoire of effector caspases. However, it was unclear whether mitochondria are involved in TNF-α-induced apoptosis, or whether activation of effector caspases is due to direct cleavage by caspase 8.

Cytochrome c release from mitochondria can be followed by immunofluorescent staining with cytochrome c-specific antibodies, and we used this method to assess whether cytochrome c release occurs when apoptosis is triggered by TNF-α. In healthy cells, cytochrome c localizes in mitochondria, and immunofluorescent staining with anti-cytochrome c antibodies produces a bright granular pattern (Fig. 3B ⇓ , bottom panel). When LNCaP cells were treated with either LY294002 or TNF-α, cytochrome c staining in some cells became uniform and less intense, which is indicative of cytochrome c release from mitochondria (designated with arrows). Cells in which cytochrome c was released from mitochondria were also stained positive by antibodies recognizing active caspase 3 (Fig. 3B ⇓ , left panel). Similar changes were observed in other cell types releasing cytochrome c from mitochondria (31) . Cytochrome c release from mitochondria in LNCaP cells undergoing apoptosis after treatment with LY294002 or lovostatin was previously shown also by subcellular fractionation (10 , 32) .

In summary, TNF-α induced the activation of caspase 8, cleavage of BID, and cytochrome c release in LNCaP cells. Thus, LNCaP cells can use the type II pathway for TNF-α-induced apoptosis.

BID Cleavage Is on the Main Pathway for TNF-α-induced Apoptosis.

Cleaved BID has been identified as the caspase 8 substrate responsible for cytochrome c release from mitochondria in cells treated with TNF-α (20 , 33) . A mechanism explaining activation of BID by caspase 8 cleavage was suggested recently (34) . In the full-length molecule, the NH₂ terminus of BID interacts with the COOH terminus and prevents association with mitochondria and subsequent apoptosis. Cleavage after Asp-59 separates the NH₂terminal part of BID from the COOH-terminal part and removes such inhibition. The larger COOH-terminal fragment then translocates to the mitochondria and triggers the release of cytochrome c and subsequent activation of caspase 9 and effector caspases.

To determine the effects of BID cleavage on apoptosis, LNCaP cells were transfected with full-length BID or C-BID mixed with GFP cDNAs in a 10:1 ratio. This method allows us to achieve expression of the protein of interest in all GFP-positive cells. Cotransfection with full-length BID resulted in approximately 20% GFP-positive cells. In the absence of proapoptotic stimuli, less then 5% of cells coexpressing BID and GFP had fragmented nuclei or altered morphology (Fig. 4A ⇓ , bottom panel; Fig. 4C ⇓ , control; see also Figs. 6 ⇓ and 8 ⇓ ). Cotransfection with C-BID yielded significantly fewer GFP-positive cells (<1%) than cotransfection with full-length BID. Most of the cells coexpressing GFP and C-BID displayed apoptotic morphology (Fig. 4A ⇓ , top panel). Treatment with ZVAD (an inhibitor of caspases with broad specificity) immediately after transfection prevented morphological changes and increased both the number of GFP-positive cells and the level of C-BID expression (Fig. 4, A and B) ⇓ . Thus, expression of C-BID was sufficient to induce apoptosis in LNCaP cells. When the N-BID fragment was coexpressed with full-length BID in higher amounts, it inhibited apoptosis induced by TNF-α (Fig. 4, C and D) ⇓ . These data are consistent with the idea that cleavage of BID contributes to apoptosis induced by TNF-α.

BID can be cleaved not only by caspase 8 but also by caspase 3, with somewhat lower efficacy (35) . Potentially, cleavage of BID by effector caspases can be used as an amplification loop for apoptosis induced via the mitochondrial pathway: effector caspases will cleave BID and trigger additional release of cytochrome c and further activation of caspases. This may not be evident, given the low levels of BID expression in LNCaP cells (Fig. 3A ⇓ , Fig. 5A ⇓ ). However, in LNCaP cells ectopically expressing BID, LY294002 can induce BID cleavage. EGF delayed cleavage of BID in LY294002-treated cells but not in cells in which apoptosis was induced by TNF-α (Fig. 5B) ⇓ . Coexpression of DN mutant of caspase 9 with BID inhibited LY294002- but not TNF-α-induced cleavage of BID (Fig. 5C) ⇓ . This is consistent with the idea that in cells treated with LY294002, we observe “retrograde” BID cleavage that is caspase 9 dependent.

When BID was transiently expressed in LNCaP cells, abundant TNF-α- and LY294002-induced apoptosis was observed 3 h after treatment, significantly earlier than in cells transfected with empty vector. As in LNCaP cells with endogenous levels of BID, EGF was able to inhibit apoptosis induced by LY294002 but not apoptosis induced by TNF-α (Fig. 6) ⇓ .

Altogether, these data are in agreement with the model that BID cleavage is on the main route of TNF-α-induced apoptosis, and this pathway is not accessible to survival signals induced by EGF or Akt. When apoptosis is induced by LY294002, BID cleavage occurs downstream of effector caspases and participates in the amplification of cytochrome c release (see the diagram in Fig. 7 ⇓ ).

Caspase 9 Is Involved in TNF-α-induced Apoptosis.

BID has been identified previously (20 , 33) as a factor inducing cytochrome c release and activation of caspases 9 and 3 in cells treated with death receptor agonists. We addressed the role of caspase 9 in apoptosis induced via different proapoptotic pathways in LNCaP cells. Expression of proteolytically inactive caspase 9 inhibited apoptosis induced by either LY294002 or TNF-α, indicating that caspase 9 activation contributes to apoptosis induced via death receptor and mitochondrial pathways (Fig. 8A) ⇓ . Note that inhibition of TNF-α-induced apoptosis was somewhat weaker than that of apoptosis induced by LY294002. It is likely that direct activation of effector caspases by caspase 8 also takes place and that TNF-α uses both type I and type II signaling to induce apoptosis in LNCaP cells.

When inactive caspase 9 was coexpressed with BID, the inhibition of TNF-α-induced apoptosis was as potent as that observed for LY294002-induced apoptosis (Fig. 8B) ⇓ . Therefore, when TNF-α-induced apoptosis is transmitted predominantly via BID cleavage, caspase 9 activation becomes essential.

These data indicate that in LNCaP cells that overexpress BID, mitochondrial and death receptor proapoptotic signals converge upstream of caspase 9 activation. Apparently, the difference between the mechanisms by which LY294002 and TNF-α induce apoptosis lies in the mechanism they use to trigger the release of mitochondrial factors activating caspase 9. TNF-α uses BID cleavage, whereas the mechanism used by LY294002 is not yet clear. The inability of EGF to inhibit TNF-α-induced apoptosis in the context of BID overexpression leads us to the conclusion that cleavage of BID, BID-induced release of mitochondrial factors, and activation of caspase 9 and effector caspases cannot be regulated by antiapoptotic signals induced by EGF. Therefore, targets of EGF-induced survival signals are localized upstream of the mitochondria, perhaps at the level of Bcl-2 proteins.

DISCUSSION

TNF-α Induces Cytochrome c Release and Bypasses Multiple Survival Signals.

When we observed that EGF is unable to protect cells from apoptosis induced by TNF-α, we were prompted to investigate how apoptosis is propagated in TNF-α-treated cells. Such an investigation was expected to help us determine the subset of proapoptotic signals targeted by EGF-induced survival signals. Therefore, it was important to determine whether proapoptotic signals induced by TNF-α and LY294002 use a completely different subset of molecules or whether their signals converge and, if so, on what level.

We found that both TNF-α and LY294002 trigger cytochrome c release from mitochondria, although the mechanisms they use may be different. Treatment with TNF-α results in cleavage of BID, and BID cleavage is sufficient to induce apoptosis in LNCaP cells. Also, expression of inactive caspase 9 can inhibit apoptosis induced by TNF-α. Because the effect of inactive caspase 9 on TNF-α-induced apoptosis was less potent than the effect on LY294002-induced apoptosis, it is most likely that both type I and type II signals coexist in these cells. However, when BID was ectopically expressed, TNF-α-induced signaling became completely inhibited by expression of inactive caspase 9, and thus, type II signaling prevailed. In this context, apoptosis induced by LY294002 and TNF-α converges at the level of mitochondria.

If survival signaling induced by EGF and other antiapoptotic agonists governed retention of cytochrome c within mitochondria or formation of the apoptosome complex activating caspase 9 and effector caspases, then propagation of TNF-α-induced apoptosis would be inhibited downstream of BID cleavage. Because EGF and other antiapoptotic agonists did not inhibit TNF-α-induced apoptosis in BID overexpressors, we concluded that targets of survival signals are upstream of cytochrome c release, most likely among proteins of the Bcl-2 family (Fig. 7) ⇓ .

Besides bypassing extracellullar survival factors, TNF-α can also induce apoptosis when Akt is constitutively active. This finding was unexpected because of reports from other groups who observed that apoptosis induced via death receptors can be inhibited by signaling via Akt (36 , 37) . One possible explanation for this discrepancy is that under the experimental conditions used in our studies (the presence of CX), only effects of posttranslational events are analyzed, and thus survival effects due to transcription or translation (e.g., activation of nuclear factor κB; Ref. 38 ) are not detected.

Although experiments with CX may be viewed as somehow artificial, agents that inhibit protein synthesis are widely used to study apoptosis induced by death receptors in tissue culture cells including prostate cancer cells (39 , 40) . This experimental paradigm allows us to analyze the interaction between proapoptotic and survival signals induced by EGF that operate solely by posttranslational modifications. Treatment with CX did not induce significant apoptosis by itself but enhanced apoptosis induced by LY294002 and TNF-α.

Role of BID and Caspase 9 in Propagation of Apoptosis in LNCaP Cells.

LNCaP cells ectopically expressing BID did not manifest increased apoptosis even when kept in serum-free conditions. However, when treated with LY294002 or TNF-α, they underwent apoptosis much more rapidly than parental cells, suggesting that BID can also be involved in the mitochondrial apoptosis pathway. We found that cleavage of ectopically expressed BID occurs in LY294002-treated cells. This is consistent with the previously reported observation that BID can be a substrate of effector caspases (35) . As a consequence, cleavage of BID can enhance activation of effector caspases by creating the amplification loop caspase 3→BID→cytochrome c→caspase 9→caspase 3 (see Fig. 7 ⇓ ). It is also possible that caspase 8 is cleaved and activated by effector caspases. In this case, the amplification loop would include caspase 8 (41 , 42) . In any case, BID appeared to be a limiting factor determining the sensitivity to apoptosis in prostate cancer cells. It would be of interest to see whether expression of BID is decreased in advanced prostate cancer.

Another observation made in the course of this work is on the role of caspase 9 in propagation of apoptosis in LNCaP cells. Recently, reports describing inhibition of apoptosis by alternatively spliced inactive variants of caspase 9 have been published (43 , 44) . In conjunction with our data on the ability of inactive caspase 9 to inhibit apoptosis induced via both mitochondrial and death receptor pathways, it would be interesting to test whether expression of inhibitory forms of caspase 9 contribute to pathogenesis of prostate cancer.

Survival Signaling and Prostate Cancer Therapy.

In this study, we made an observation that in prostate cancer LNCaP cells, survival signals can inhibit apoptosis induced via the mitochondrial pathway but are ineffective if apoptosis is induced by activation of death receptors. Analysis of proapoptotic signaling induced by LY294002 and TNF-α allowed us to position the site of action of EGF-induced survival signals upstream of mitochondria.

Our results suggest that therapeutic agents inducing apoptosis via the mitochondrial pathway will be much less effective in LNCaP cells than agents operating by activation of death receptor signaling. It would be of great interest to see whether this observation can be generalized to other cancer-derived cell lines. Indeed, recent reports on the high sensitivity of several cancer cell lines to tumor necrosis factor-related apoptosis-inducing ligand-induced apoptosis (45) and increased apoptosis in LNCaP cells treated by a combination of γ-irradiation and TNF-α (46) support the idea that signaling via death receptor pathways can be a promising avenue for design of effective therapy against advanced prostate cancer and perhaps other malignancies.