Alternative programs of cell death in developing retinal tissue.

We examined cell death in developing retinal tissue, following inhibition of protein synthesis, which kills undifferentiated post-mitotic cells. Ultrastructural features were found of both apoptosis and autophagy. Only approximately half of the degenerating cells were either terminal dUTP nick-end labeling (TUNEL)-positive or reacted with antibodies specific for activated caspases-3 or -9. Bongkrekic acid completely inhibited any appearance of cell death, whereas inhibitors of autophagy, caspases-9 or -3, prevented only TUNEL-positive cell death. Interestingly, inhibition of caspase-6 blocked TUNEL-negative cell death. Simultaneous inhibition of caspases-9 and -6 prevented cell death almost completely, but degeneration dependent on autophagy/caspase-9 still occurred under inhibition of both caspases-3 and -6. Thus, inhibition of protein synthesis induces in the developing retina various post-translational, mitochondria-dependent pathways of cell death. Autophagy precedes sequential activation of caspases-9 and -3, and DNA fragmentation, whereas, in parallel, caspase-6 leads to a TUNEL-negative form of cell death. Additional mechanisms of cell death may be engaged upon selective caspase inhibition.

Programmed cell death is a major component of both normal development and disease (1)(2)(3)(4)(5)(6)(7)(8)(9) and occurs in various forms, such as apoptosis, autophagy, and others (for reviews, see Refs. 10 -12). Identification of the mode of cell death in each circumstance is important for the understanding of mechanisms, as well as for the design of intervention strategies.
Cell death with apoptotic morphology can be triggered by various stimuli, including intracellular stress and activation of membrane receptors (13). These signals engage an evolutionarily conserved and ubiquitous intracellular machinery of execution, that depends on the caspase family of cysteine-aspartate proteases (for reviews, see Refs. 14 and 15).
Three major pathways of apoptosis have been identified ac-cording to their initiator caspase: the death receptor pathway involving caspase-8 (16); the endoplasmic reticulum stress pathway attributed to activation of caspase-12 (17); and the mitochondrial pathway, in which the release of harmful proteins by mitochondria into the cytoplasm leads to activation of caspase-9 and downstream cleavage and activation of caspase-3, -7, or -6 (for reviews, see Refs. 18 and 19). The latter enzymes, known as effector caspases can, nonetheless, be also activated through mechanisms independent of caspase-9 (20 -23).
Autophagy has been identified both as a means to resist starvation and as part of cellular remodeling, as well as in the removal of anomalous cellular components that accumulate following toxic insults or during cell death (24 -27). For example, morphological signs of autophagy are observed during the continuous trimming of photoreceptor outer segments by the retinal pigment epithelium (28), which is a physiological event not associated with cell death. Notwithstanding, autophagy may lead to cell death either per se (29 -32) or associated with signs of apoptosis (33)(34)(35)(36).
Blockade of caspase-mediated apoptosis may allow a complete program of cell death to proceed based on the autophagosomal-lysosomal compartment (31,34,35), thus uncovering alternative pathways of cell death in individual cells. On the other hand, the apoptotic execution machinery is available for action in cells independent of protein synthesis (37), and currently identified components of autophagy are also associated with post-translational, rather than transcriptional control mechanisms (27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37)(38). Thus, it appears that, in both forms of cell death, either the nature of the upstream cell death-inducing signals or the metabolic state of each cell determines its mode of triggering an essentially ready-made program of cell demise.
Notwithstanding that several programs of cell death have been demonstrated in a variety of cell lines or isolated cultured cells, little is known of the relationship between apoptosis and autophagy within organized tissues, particularly during development. However, an association of autophagy and apoptosis has been shown to underlie the involution of several organs in metamorphosing insects (reviewed in Ref. 12).
Using explants of developing retina, we have previously shown that inhibition of protein synthesis triggers apoptosis restricted to a relatively homogeneous population of undifferentiated, post-mitotic cells located within the neuroblastic (proliferative) layer (39,40). The histotypical retinal explants are a favorable model to study mechanisms of cell death in a complex and highly structured tissue such as the mammalian central nervous system, because they preserve the spatial distribution of various cell types, cell communication including gap junctions, as well as the extracellular matrix (41,42), all of which affect the sensitivity to cell death (43). In addition, the retinal explant model preserves interactions of cells in various stages of differentiation, a major trait of developing complex tissues. Within this context, programmed cell death induced by inhibition of protein synthesis allows the analysis of the post-translational execution programs, independent of upstream gene expression (37).
In this study we examined the morphology and mechanisms of cell death induced by inhibition of protein synthesis in retinal tissue, with emphasis on the relationship between apoptosis and autophagy.

EXPERIMENTAL PROCEDURES
Materials-Anisomycin (Sigma), an inhibitor of protein synthesis, was used at 1 g/ml from a stock solution in Me 2 SO. 3-Methyladenine (3MA 1 ; Sigma), an inhibitor of autophagy, was used at 1-10 mM from a stock solution in culture medium. Ac-Leu-Glu-His-Asp-H (aldehyde) (Ac-LEHD-CHO), an inhibitor of caspase-9, and Ac-Val-Glu-Ile-Asp-H (aldehyde) (Ac-VEID-CHO), an inhibitor of caspase-6, were purchased from Peptide Institute, while Ac-Asp-Glu-Val-Asp-fluoromethylketone (Ac-DEVD-fmk) was purchased from Calbiochem, and were used at 10 -100 M from stock solutions in Me 2 SO. Bongkrekic acid (Calbiochem), an inhibitor of adenine nucleotide translocator, was used at 3-100 M from a stock solution in sterile water. Ac-VEID-p-nitroanilide, a substrate for caspase-6, was used at 200 M from a stock solution in Me 2 SO.
Tissue Culture-Lister hooded rats were used in this study. Retinal explants were prepared as previously described (44). The animals were killed instantaneously by decapitation at postnatal day 1, their eyes were removed, and the retinas were dissected. Fragments of ϳ1 mm 2 were cut in culture medium and placed in 25-ml tight-lidded Erlenmeyer flasks with 5 ml of Eagle's basal medium (Invitrogen) with 5% fetal calf serum (WL Immunochemicals) and 20 mM Hepes. The flasks were kept in an orbital shaker at 80 -90 revolutions/min and 37°C for 24 h. Drugs were added at the beginning of the incubation period.
Histology-The tissue was fixed with 4% paraformaldehyde in 0.1 M sodium phosphate buffer, pH 7.2, for 1 h and then infiltrated in a solution of 30% sucrose in 0.1 M sodium phosphate buffer, pH 7.2. The explants were oriented under a dissecting microscope, in an aluminum chamber filled with OCT embedding medium, and transverse sections were cut at 10-m thickness in a cryostat. The sections were either stained with neutral red or processed for immunocytochemistry or TUNEL staining.
TUNEL Staining-In situ nick-end labeling of fragmented DNA (TUNEL) was done using a FraGel kit (Oncogene).
Immunocytochemistry-Immunocytochemistry for activated caspase-9 was done using cleaved Caspase-9 (Asp-353) antibody (Cell Signaling Technology) as recommended by the manufacturer. Briefly, tissue sections were incubated with Tris-buffered saline (TBS)/Triton, washed with TBS, and then incubated with blocking solution (5.5% normal goat serum in TBS/Triton) for 1 h at room temperature. After blocking, sections were incubated with primary antibody at the suggested dilution (1:100) in 5% bovine serum albumin (BSA) in TBS, overnight at 4°C. The sections were washed with TBS/Triton and incubated with biotinylated secondary antibody (diluted in TBS plus 3% BSA; 1:200 for secondary antibody from Vectastain HRP-ABC kit, Vector Laboratories) for 1 h at room temperature. After washing with TBS/Triton, sections were incubated for 30 min with 0.6% H 2 O 2 in TBS at room temperature. Sections were washed with TBS/Triton and then incubated for 1 h with ABC reagent at room temperature. After washing with TBS, sections were incubated with diaminobenzidine for 7 min, washed with water, coverslipped in glycerol, and analyzed by light microscopy.
Immunocytochemistry for activated caspase-3 was done using the CM1 polyclonal antiserum that recognizes the p20 activated form of human, mouse, and rat caspase-3, without significant cross-reactivity with the zymogen form of 32 kDa (45). This antibody was a gift from Dr. Anu Srinivasan (Idun Pharmaceuticals), and a detailed description was provided in the publication mentioned above. Paraformaldehyde-fixed tissue sections were incubated with 3% H 2 O 2 in phosphate-buffered saline (PBS) for 30 min, followed by a PBS wash. The sections were incubated with a blocking buffer containing 2% normal goat serum, 2% BSA, 0.2% nonfat milk powder, and 0.8% Triton X-100 in PBS for 1 h at room temperature. After this period, the sections were incubated overnight at 4°C with primary antibody CM1 at 1:5000. The immunocytochemical reaction was developed with a Vectastain HRP-ABC kit, and the sections were coverslipped in glycerol and analyzed by light microscopy.
Quantification of Cell Death and Statistics-Dying cells appeared as round condensed profiles, either homogeneously stained with Neutral Red or as outstanding globose profiles under differential interference contrast microscopy, usually found isolated among healthy looking cells. To estimate the rates of cell death, either in Neutral Red-stained material or in immunocytochemistry/TUNEL-stained material, condensed profiles were counted within the neuroblastic layer (NBL) in three random fields of 0.0148 mm 2 from each of three distinct explants in each experiment. Data were expressed as means Ϯ S.E. of the replicates from several pooled experiments. The values were normalized with respect to the average rate of cell death within the NBL in anisomycin-treated explants in each experiment, which was taken as 100%. The data were subject to either one-way or two-way analysis of variance, as applicable, followed by Duncan's multiple range test, using an SPSSPC statistical software. p Ͻ 0.01 was set as criterion significance.
Transmission Electron Microscopy-After 24 h in culture in either control medium or incubated with anisomycin, retinal explants were fixed overnight in a solution containing 2.5% glutaraldehyde, 4% freshly prepared formaldehyde in 0.1 M phosphate buffer, pH 7.3. After fixation the cells were washed in PBS and post-fixed on 1% OsO 4 in cacodylate buffer plus 5 mM CaCl 2 and 0.8% potassium ferricyanide. Cells were washed, stained with uranyl acetate and lead citrate, dehydrated at 20°C in an acetone series, and infiltrated at the same temperature in Epon. Polymerization was carried out for 72 h at 37°C. Thin sections were collected on copper grids, stained with uranyl acetate and lead citrate, and examined in a JEOL 1210 transmission electron microscope.
Caspase Activity Assay-Protein samples were prepared as follows. Retinal explants cultured for 24 h in various experimental conditions (indicated in the legend) were incubated with buffer A (150 mM NaCl, 20 mM Tris, pH 7.5, 1% Triton X-100, 10 g/ml leupeptin, 5 g/ml aprotinin, 100 M phenylmethylsulfonyl fluoride) for 10 min on ice. Then buffer B was added (100 mM Hepes, pH 7.4, 150 mM NaCl, 0.2% CHAPS, 4 mM dithiothreitol), the samples were centrifuged, and the pellet was stored at Ϫ70°C. Protein samples (300 g) were incubated at 37°C for 4 h with VEID-p-nitroanilide (200 M in buffer B), and substrate hydrolysis was measured by sample absorbance at 405 nm in a Microplate Reader 3550-UV (Bio-Rad).

Inhibition of Protein Synthesis Induces Ultrastructural Features of Both Apoptosis and Autophagy in Retinal Explants-In
retinal tissue stained with neutral red, dying cells appear as round and heavily stained profiles, easily detectable among healthy cells. (Fig. 1A). Examples of low and high power views of sections of retinal tissue containing degenerating cells can be found in our previous publications (see, e.g., Refs. 40 and 46). These condensed profiles can be also detected by their globose morphology, under differential interference contrast microscopy (Fig. 2, A-C). In explants maintained in control medium, ganglion cells degenerate as a result of axotomy, whereas there is no cell death in the outer stratum of the retina. In retinal tissue treated with the inhibitor of protein synthesis anisomycin (ANI), ganglion cells degeneration is blocked, whereas cell death is induced in the NBL. Although ANI is known to activate mitogen-activated protein kinases at low concentrations, cell death in the NBL is induced only at concentrations of anisomycin that significantly block protein synthesis (46). 2 The paradoxical effect of anisomycin was described previously by Rehen and co-workers (40), who also demonstrated that the cells sensitive to ANI-induced cell death in the developing 1 The abbreviations used are: 3MA, 3-methyladenine; Ac-LEHD-CHO or LEHD, Ac-Leu-Glu-His-Asp-H; Ac-VEID-CHO or VEID, Ac-Val-Glu-Ile-Asp-H; ANI, anisomycin; APAF-1, apoptotic protease activating factor-1; NBL, neuroblastic layer; TUNEL, terminal dUTP nick-end labeling; Z-DEVD-fmk or DEVD, Ac-Asp-Glu-Val-Asp-fluoromethylketone; TBS, Tris-buffered saline; BSA, bovine serum albumin; PBS, phosphate-buffered saline; TNF, tumor necrosis factor; CHAPS, 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonic acid.
retina are undifferentiated post-mitotic cells (39). We did not observe massive degenerative areas, and dead cells were always found surrounded by healthy-looking profiles, showing that anisomycin does not induce tissue necrosis. Ultrastructural analysis of the NBL in control explants from the retinas of neonatal rats showed the presence mostly of elongated nuclei containing granular and relatively diffuse chromatin, surrounded by very small amounts of cytoplasm, typical of immature cells of the proliferative neuroepithelium (Fig. 1B). No attempt was made to distinguish between the proliferating and post-mitotic cells known to reside within this layer (39). The massive predominance of vertically elongated nuclei within the NBL is probably associated with the ongoing vertical migration of the nuclei of cells within this layer. Proliferating cell nuclei move to and fro as a result of interkinetic nuclear migration, whereas post-mitotic cells usually translocate vertically toward their definitive positions in the emerging retinal layers. Occasional larger, round nuclei are found toward the external margin of the NBL, probably corresponding to a much smaller population of early differentiating horizontal cells (data not shown).
In tissue treated with anisomycin, nuclei showing various degrees of chromatin condensation were typical of apoptosis (Fig. 1C). Around the presumptive early apoptotic profiles occasional organelles appeared preserved (data not shown), whereas the whole cell bodies rounded up around the heavily condensed late apoptotic nuclei. Typical autophagic profiles were also frequently found within the cytoplasm of cells in the NBL of retinal tissue treated with anisomycin ( Fig. 1

, D-F).
Either portions of the cytosol or whole organelles were found within double-or multiple-layered vacuoles (Fig. 1D). Occasionally, clusters of mitochondrial profiles were found surrounded by a double-membrane vacuole (Fig. 1, E and F). These ultrastructural features are consistent with the induction by anisomycin of both apoptosis and autophagy in retinal tissue.
Anisomycin Induces Cell Death Only in Part through Activation of Caspases-9 and -3 and DNA Fragmentation-Both a DNA ladder and TUNEL staining of degenerating profiles within the NBL have been previously shown in retinal tissue treated with anisomycin (40). However, only a fraction of the globose, degenerating profiles observed under differential interference contrast were stained by the TUNEL procedure. 2 We systematically quantified TUNEL labeling in the NBL of retinal explants treated with ANI and found that 57 Ϯ 5.6% (mean Ϯ S.E.) of the pyknotic profiles were positive (Fig. 2, A  and D). These results suggested that anisomycin-induced cell death may be either TUNEL-positive or -negative.
We then examined retinal explants treated with anisomycin using the CM1 antibody raised against the activated form of caspase-3. Only 55 Ϯ 6.3% of the condensed profiles were immunoreactive (Fig. 2, B and E), supporting the hypothesis that anisomycin induces two distinct pathways of cell death, one of which seems to be dependent on caspase-3 activation.
Furthermore, only 50 Ϯ 2.2% of condensed cells were immu- noreactive with an antibody to activated caspase-9 (Fig. 2, C and F), consistent with the hypothesis that caspase-9 is the protease responsible for the activation of caspase-3 in the retina. Therefore, the quantitative data indicated that a cascade of activation of caspase-9, caspase-3, and DNA fragmentation leading to TUNEL staining appears to be associated with only a fraction of the cell death induced in retinal tissue by inhibition of protein synthesis. It is unlikely that these findings can be explained by a trivial failure of the staining methods to attain all cells, both because of the similar fractions of cells stained with three distinct procedures and because of the consistent effects of selective caspase inhibitors described below.
Mitochondria Are Involved in Cell Death Induced by Anisomycin-Activation of caspase-9 is dependent on the assembly of the mitochondrial apoptosome, containing procaspase-9, APAF-1, dATP, and cytochrome c. The release of cytochrome c is often associated with the opening of a permeability transition pore in the outer membrane of the mitochondria. To test whether mitochondria participate in cell death induced by ANI, we used bongkrekic acid, an inhibitor of the adenine nucleotide translocator, which is one of the components of permeability transition pore. Bongkrekic acid completely inhibited cell death induced by ANI (Fig. 3), showing that cell death induced by ANI is dependent on mitochondrial commitment.
Inhibitors of Caspases-9, -3, and -6 and Autophagy Block in Part Cell Death Induced by Anisomycin-To test for the role of various caspases in retinal cell death, we treated retinal explants in the presence of ANI together with each one of various peptides that inhibit caspase-9, caspase-3, or caspase-6. LEHD, an inhibitor of caspase-9, partially blocked cell death induced by ANI (fig 4A). LEHD concentrations up to 300 M had no additional effect, reaching a maximum near 60% inhibition. This is consistent with the immunoreactivity for activated caspase-9, found in roughly half of the degenerating profiles (Fig. 2, C and F). DEVD, an inhibitor of caspase-3, partially blocked cell death induced by ANI (Fig. 4B), also in agreement with immunocytochemical data for activated caspase-3 (Fig. 2,  B and E). VEID, an inhibitor of caspase-6, also partially blocked cell death induced by ANI (Fig. 4C). These results show the participation of at least one initiator and two effector caspases in cell death induced by ANI in the retina. However, none of these appears to be responsible for all the cell death that follows inhibition of protein synthesis.
Consistent with electron microscopy (Fig. 1), 3MA, an inhibitor of autophagy (34,47), decreased the rate of cell death in ANI-treated explants (Fig. 4D). The inhibitor effect of 3MA upon ANI-induced cell death reached a maximum at a concentration of 3 mM, above which the drug alone induced cell degeneration in the NBL. These data are consistent with the hypothesis that ANI-induced cell death in the retina involves autophagy.
Autophagy and Apoptosis Are Stages of the Same Cell Death Pathway-To test the hypothesis that autophagy and apoptosis share an intracellular signaling pathway in retinal tissue, we treated retinal explants with ANI plus either 3MA or LEHD, or both together (Fig. 5A). The latter did not have additive effects upon cell death when compared with each inhibitor separately. This suggests that autophagy and caspase-9 participate in the same cell death pathway.
Retinal explants treated with ANI and 3MA were immunostained with antibodies to activated caspases-9 and -3. Treatment with 3MA reduced markedly the number of condensed profiles stained for activated caspase-9 (Fig. 5B), as well as for caspase-3 (Fig. 5C). As expected, treatment with the caspase-9 inhibitor LEHD reduced the number of condensed profiles stained for activated caspase-3 (Fig. 5D), and the peptides LEHD and DEVD had no additive effects when compared with each inhibitor alone (data not shown).
We also stained explants treated with 3MA by the TUNEL procedure. The number of TUNEL-stained condensed profiles decreased with increasing concentrations of 3MA, whereas the number of TUNEL-negative cells was unchanged (Fig. 5E). These results altogether indicate that ANI induces a cell death pathway involving an autophagic phase followed by sequential activation of caspases-9 and -3 and fragmentation of DNA.

Caspase-9 and Caspase-6 Mediate Parallel Pathways of Cell
Death in Retinal Tissue-To test whether caspases-9 and -6 are in the same pathway, we treated retinal explants with ANI plus either LEHD or VEID, or both together (Fig. 6A). Coincubation with both inhibitors further protected cells from ANI-induced cell death, compared with LEHD or VEID alone, suggesting that caspases-9 and -6 are activated in parallel pathways. In addition, VEID had no effect upon immunoreactivity for either activated caspase-9 or -3 (Fig. 5, B and C). We also measured caspase-6 activity in retinal explants treated with either LEHD or VEID (Fig. 6B). Whereas VEID reduced caspase-6 activity to control levels, LEHD had no effect, showing that activation of caspase-6 in retinal tissue treated with anisomycin is not dependent on caspase-9.
Treatment of retinal explants with both 3MA and VEID (Fig.  6C) also led to a synergic protective effect, consistent with the hypothesis that caspase-6 is activated in a distinct pathway from autophagy/caspase-9/caspase-3.
Notwithstanding the clear synergic effect of the combined treatments with the caspase-6 inhibitor plus the inhibitors of either caspase-9 or autophagy, a residual number of condensed profiles remained significantly above control levels in both circumstances (Fig. 6, A and C).

Caspase-9 Induces an Alternative Pathway of Cell Death in
Retinal Tissue-The results described above demonstrate that the effector caspases-3 and -6 participate in two distinct cell death pathways induced by ANI. Thus, we expected that DEVD and VEID would have a synergic effect upon cell death in retinal cells. Surprisingly, co-incubation of retinal explants with DEVD plus VEID had no further effect upon anisomycininduced cell death, when compared with each peptide separately (Fig. 7A). This result raised the possibility that an alternative cell death pathway may be activated when both effector caspases are blocked.
To test whether this pathway depends on autophagy or caspase-9, we treated explants with various combinations of VEID, DEVD, 3MA, and LEHD. Either the combination of DEVD/VEID/3MA or VEID/DEVD/LEHD further protected retinal cells from cell death induced by ANI compared with the combination of DEVD plus VEID (Fig. 7B). This shows that an alternative cell death pathway, induced when effector caspases-3 and -6 are inhibited, follows autophagy and activation of caspase-9. However, this synergic effect did not lead to complete blockade of ANI-induced cell death, and a significant residual number of condensed profiles remained in the NBL after triple inhibition (Fig. 7B). DISCUSSION The major findings of this study were as follows. 1) Inhibition of protein synthesis induces ultrastructural signs of 3) A caspase-9-dependent pathway is preceded by autophagy, and followed by activation of caspase-3 and TUNEL-positive DNA fragmentation, whereas a parallel caspase-6-dependent pathway leads to TUNELnegative cell death. 4) Additional alternative pathways of cell death may be engaged upon inhibition of both effector caspases-3 and -6.
Retinal Cell Death Induced by Inhibition of Protein Synthesis-Although dependence on both transcription and protein synthesis is a frequent component of programmed cell death (48 -52), it has been repeatedly shown that apoptosis can occur either independent of, or as a consequence of inhibition of protein synthesis (40,53,54). Recently, it was shown that cell death induced by inhibitors of protein synthesis in a T-lymphocyte cell line is associated with cytochrome c release and activation of caspase-9 and can be blocked by Bcl-2 overexpression (37). These data demonstrate that most, if not all, of the machinery of execution of the cell death program is constitutive and is activated post-translationally on demand after a noxious stimulus. Thus, the use of inhibitors of protein synthesis as inducers of cell death allows the isolated analysis of the execution pathways of cell death. Furthermore, analysis of cell death induced by translation inhibitors may be of clinical relevance, because inhibition of protein synthesis occurs in neurons during ischemia, immediately upon reperfusion and persists particularly in regions known to suffer extensive cell death (55)(56)(57). It should be noted that, within the time frame examined in our model, cell death following inhibition of protein synthesis is restricted to a specific population of cells (39), and therefore does not represent a trivial metabolic effect.
In our model of retinal tissue culture, cell death following inhibition of protein synthesis was accompanied by chromatin condensation, in a pattern similar to classically defined apoptosis (58). The cells in the neuroblastic layer are immature, and their scanty cytoplasm makes it difficult the simultaneous study of both nuclear and cytoplasmic changes in the same cell. However, preserved organelles were found in the cytoplasm around nuclear profiles showing early signs of condensation. In addition, typical autophagic vacuoles were detected within the neuroblastic layer after treatment with anisomycin, and also in apical processes of cells from the neuroblastic layer. Although it has been shown that neuroepithelial cells have the capacity to phagocytose apoptotic bodies in the early developing retina (59), the double-or multiple-membrane vacuoles often filled with numerous profiles of a single type of organelle and devoid of cytoplasmic or nuclear fragments are typical of autophagy. These data support the hypothesis that both apoptosis and autophagy occur in retinal tissue as a consequence of inhibition of protein synthesis.
Pathways of Caspase-dependent Cell Death in the Retinal Tissue-Caspases have been implicated in retinal cell degeneration both under experimental conditions, and in the course of various pathologies, such as glaucoma, ischemic insults, and stress-mediated degeneration of photoreceptors (60 -67). However, the quoted studies tested individual members of the caspase family, without examining either the sequence of activation or interplay of these proteases. The present study represents the first attempt to systematically examine the sequential activation of caspases within retinal tissue.
Inhibition by bongkrekic acid shows that anisomycin-induced cell death depends on mitochondrial commitment. The exact relationship between the opening of permeability transition pores in the mitochondrial membrane, and the release of various harmful proteins, including cytochrome c, is still unclear. Nevertheless, bongkrekic acid was the only drug that completely abolished cell death following treatment with the protein synthesis inhibitor.
Following the mitochondrial step, activation of caspases was found to be required for anisomycin-induced cell death. Peptide inhibitors are not completely specific for each caspase, despite their significant selectivity (68,69). Nevertheless, cross-inhibition of the relevant caspases among the various peptides seems to be of little significance in our experiments. For example, increased concentration of the caspase-9 inhibitor LEHD did not lead to the same effect as addition of the caspase-6 inhibitor VEID, and their effects were opposite upon either TUNELpositive or TUNEL-negative profiles. In addition, the two peptides had distinct effects upon the biochemical measurements of caspase activity. Furthermore, either DEVD or LEHD had distinct effects upon cell death when combined with VEID, and either DEVD or VEID also had distinct effects when combined with LEHD. Therefore, together with both the immunohistochemistry and TUNEL procedure, the pharmacological approach showed consistently that a caspase-9/caspase-3/ TUNEL-positive DNA fragmentation sequence was dependent on an autophagic step, blocked by 3-methyladenine. A parallel pathway was responsible for simultaneous TUNEL-negative cell death, dependent on the activation of caspase-6 ( Fig. 8).
Whereas the mitochondria-dependent, apoptosome-mediated sequential activation of caspase-9 and caspase-3 is a common pathway of apoptosis (for review, see Ref. 19), caspase-3 can also be directly activated by caspases-8 and -6 (19 -70). A role for caspase-8 was reported for cell death induced by an antibody directed to Hsp27 in a retina-derived cell line (71). However, in the present study the abolishment of immunoreactivity for activated caspase-3 in explants treated with the inhibitor of caspase-9, apparently rules out the involvement of caspase-8 as the activator of caspase-3 in anisomycin-induced cell death in the retina. Caspase-6, in turn, partially blocked anisomycininduced cell death without preventing caspase-3 activation, as shown immunohistochemically.
Caspase-6 can be activated through pathways either dependent or independent of caspase-9, leading to apoptotic cell death (20,70,72,73). In our experiments, however, the effects of inhibitors of either caspase-9 or -6 were distinct. The mechanisms of activation of caspase-6 remain to be clarified, but it has not been shown to immediately follow mitochondrial events, and it is unlikely to occur through oligomerizationinduced activation, because caspase-6 does not contain the domains required for autoprocessing found in initiator caspases (for review, see Ref. 74).
A candidate protease to activate caspase-6 is caspase-2 (75,76). The activation of caspase-2 may precede mitochondrial commitment, therefore also preceding the activation of caspases-9 and -3 (77,78). On the other hand, caspase-2 is not necessary for the cell death of either thymocytes or dorsal root ganglion neurons, although its cleavage in these cells depends on both APAF-1 and caspase-9 (75).
The precise mechanism of caspase-2 activation is still unknown, except in the cases where it is activated by RAIDD (receptor-interacting protein-associated ICH-1 homologous protein with death domain) (79). However, the presence of this protease within mitochondria and its release upon apoptotic insults (80) allow the hypothesis that caspase-2 may be involved in the activation of caspase-6 after mitochondrial commitment in the retina. Indeed, preliminary experiments showed both activation of caspase-2, and blockade of cell death with a caspase-2 inhibitor peptide in retinal explants treated with anisomycin. 2 An alternative possibility is that caspase-6 is activated by serine proteases. In fact, preliminary experiments showed that aprotinin, a generic inhibitor of serine proteases, blocks roughly 50% of the cell death induced by anisomycin in retinal explants. 2 Downstream of caspase-6, one possible pathway leading to TUNEL-negative cell death involves the activation of L-DNase II, which plays a role in degenerative events in the visual system, including the retina (81,82). L-DNase II cleaves DNA in single and double ends according to the ionic conditions, but always produces 3Ј-phosphate ends, which are not accessible to the enzyme TdT, and therefore preclude TUNEL staining. 3 Further work will be needed to test for a link of caspase-6 with L-DNase II.
The lack of synergy between inhibitors of caspases-3 and -6, together with the results of triple inhibition of both caspase-3 and -6 plus either caspase-9 or autophagy, suggest that autophagy/caspase-9 may induce an alternative cell death program apart from activation of caspase-3. This alternative route appears to be inhibited when either of the effector caspases-3 or -6 are activated (Fig. 8). However, it is also conceivable that a caspase-9-dependent alternative route may proceed at a slower pace in the absence of both caspase-3 and caspase-6 activities, albeit leading to a similar morphology of the degenerating cells at 24 h of incubation. The alternative pathway cannot be attributed to caspase-7, because the inhibitor DEVD also blocks this caspase (83). Caspase-9 substrates other than effector caspases-3 and -7 have already been described, such as vimentin (84). Activated caspase-9 was also found within the nucleus of dying cells, raising the possibility of additional nuclear substrates (85).
It should be noted that simultaneous treatment with inhibitors of caspases-9, -3, and -6 did not completely block cell death induced by anisomycin, distinct from the inhibition of mitochondrial transition permeability pore with bongkrekic acid. It is possible that the remaining cell death found upon simultaneous inhibition of caspases-9, -3, and -6 is the result of proapoptotic factors released from mitochondria, such as either apoptosis-inducing factor (86) or endonuclease G (87), which do not require caspases to lead to cell death (Fig. 8).
Although our experiments do not rule out that distinct programs may apply, in part, to distinct cell populations, it should be emphasized that the cells sensitive to anisomycin-induced cell death are a relatively homogeneous population of undifferentiated cells within 3-4 days following the last mitosis, at least 80% of which are bound to differentiate into photoreceptors (39). Therefore, it is likely that several, if not all, of the alternative pathways described here may be triggered in any one cell among this population.
Autophagy and Apoptosis in Retinal Tissue-Autophagy has been implicated in several neurodegenerative disorders, such as Alzheimer, Parkinson, Huntington, and prion diseases (88 -93). Light-induced stress induces ultrastructural changes in rod photoreceptors, including proliferation of autophagic bodies in the inner segments of these cells (94), which suggests a role for autophagy in retinal degeneration. In the present study, we detected ultrastructural hallmarks of autophagy in the retina, and, using the inhibitor 3-methyladenine, we partially inhibited anisomycin-induced cell death. Furthermore, we demonstrated that an autophagic step is required before sequential activation of caspases-9 and-3.
Bursch and collaborators (29) have shown that the MCF-7 breast carcinoma cell line, which does not express caspase-3 (95), undergoes autophagic cell death upon treatment with tamoxifen. Recent work in cell lines showed that apoptotic and autophagic cell death are, respectively, accompanied by either degradation or rearrangement of cytoskeletal proteins (30). These data support the concept that autophagy is a separate form of programmed cell death distinct from apoptosis.
In particular, the role of cathepsins in the execution of autophagic cell death has been emphasized recently (12,31,35). For example, serum deprivation of PC12 cells leads to ultrastructural signs of autophagy, accompanied by changes in the activity of lysosomal proteases, particularly cathepsins B and D (35). However, the roles of individual lysosomal proteases in 3 A. Torriglia, personal communication. In addition to the parallel pathways mediated either by caspase-3 or by caspase-6, the diagram indicates that an alternative pathway dependent on autophagy/caspase-9 may be inhibited by either caspase-3 or -6, because it appears only when peptide inhibitors for both caspases are used. This may, however, be the result of a slower pace of the alternative route of cell death. A minor component derived from mitochondria without involvement of caspases -9, -3, or -6 is suggested by residual cell death in the presence of inhibitors for all these three caspases.
the execution of autophagy appear to depend on either cell type or circumstances. For example, although in nerve growth factor-deprived PC12 cells cathepsin D favors, and cathepsin B counteracts, cell death in the presence of a caspase-3 inhibitor (see Ref. 35 for review), activation of cathepsin B appears to be associated with the promotion of cell death in both tumor necrosis factor (TNF)-induced tumor cell death (96), and in involuting salivary glands of metamorphosing insects (12). Further work will be necessary to define the role of lysosomal proteases in autophagy in developing retinal tissue.
In contrast, other studies have shown that apoptosis and autophagy may not be independent modes of cell death. Cell death induced by 5-fluoruracil in keratinocytes is accompanied by features of autophagy, such as cytoplasm vacuolization and chromatin detachment from the nuclear membrane, as well as of apoptosis, such as TUNEL staining and a decrease in size and increase in granularity observed by cytometric analysis (36). The authors, however, argued that TUNEL is not specific for apoptosis, assuming that in this case TUNEL-positive cells are dying by autophagy. Following axotomy of peripheral nerves in adult rats, oligodendrocytes showed both ultrastructural features of apoptosis as well as organelles engulfed by double-membrane vacuoles in the cytoplasm, typical of autophagy (97). These data demonstrate that mixed cell death features may be more common than predicted by previous studies.
Jia and co-workers (33) showed that TNF-␣ induces cell death in an acute T-lymphoblastic leukemia cell line, with an apoptotic pattern preceded by autophagy. 3-Methyl-adenine, that inhibits the formation of autophagosomes, inhibited the cytolysis and DNA fragmentation induced by TNF-␣. However, inhibition of fusion of lysosomes with autophagosomes by asparagine did not block TNF-␣-induced apoptosis, and amino acid and protein deprivation enhanced TNF-␣-induced autophagy but not apoptosis. In this case, early stages of autophagy are required for, but do not necessarily result in, TNF-␣-induced apoptosis.
In sympathetic neurons from the superior cervical ganglia, growth factor deprivation induced autophagy, and treatment of these cells with 3-methyladenine decreased the rate of apoptosis (34), consistent with the hypothesis that autophagy may be a necessary step toward apoptosis in some cells. In contrast, the autophagy inhibitor 3-methyladenine increased the sensitivity of HT-29 colon cancer cells to apoptosis induced by sulindac sulfide, a cyclooxygenase inhibitor. Mutants with a low rate of autophagy were more sensitive to apoptosis than parental HT-29 cells, and the rate of cytochrome c release was higher in mutant cells than in HT-29 cells, suggesting that autophagy may delay apoptosis by sequestering mitochondrial death-promoting factors such as cytochrome c (98). In this context, autophagy may represent an attempt of the cell to recover from a noxious stimulus, rather than a necessary stage of execution of cell death.
Therefore, in some cases cells can die by autophagy, independent of caspase activation or other features of apoptosis, and both types of cell death may represent the ends of a wide spectrum of physiological cell death (99). In addition, lysosomal destabilization followed by the release of cathepsins may lead to mitochondria-dependent, caspase-mediated apoptosis (100,101), although it is not clear how this mechanism relates to conventional autophagy pathways. However, there are circumstances where apoptosis and autophagy may pertain to the same pathway of programmed death. In developing retinal tissue, our data indicate that a post-mitochondrial autophagic step is necessary for one pathway of cell death, similar to growth factor-deprived sympathetic neurons (34), whereas a parallel pathway, also dependent on mitochondrial commit-ment, does not require the autophagy step. In addition, alternative routes of cell death are programmed and ready for use upon demand in retinal tissue.
Our data add to a growing body of evidence for multiple intrinsic pathways of programmed cell death. The alternative mechanisms underscore the need to identify the specific pathways and critical components of cell death in each case, and make it unlikely that any approach based on a single execution pathway may lead to full-proof control of developmental cell death or therapeutics for degenerative diseases.