Multiple pathways of apoptosis in PC12 cells. CrmA inhibits apoptosis induced by beta-amyloid.

Stable transfectants of PC12 cells expressing bcl-2 or crmA were generated and tested for their susceptibility to various apoptotic insults. Bcl-2 expression conferred resistance to apoptosis induced by staurosporine and by oxidative insults including hydrogen peroxide and peroxynitrite, but was less effective in inhibition of activation-induced programmed cell death induced by concanavalin A. Concanavalin A-induced apoptosis was abated, however, in cells expressing very high levels of bcl-2. In contrast, cells expressing crmA were protected from concanavalin A-induced apoptosis, but were as susceptible as control cells to apoptosis induced by staurosporine and oxidative insults. Therefore, at least two apoptotic pathways in PC12 cells can be discerned by their differential sensitivity to blockade by bcl-2 and crmA. The ability of beta-amyloid (Abeta) to induce apoptosis in these cells was assessed. CrmA transfectants were protected from apoptosis induced by Abeta1-42, but only cells expressing very high levels of bcl-2 were similarly protected. These results suggest that the apoptotic pathway activated by Abeta1-42 in PC12 cells can be differentiated from the apoptotic pathway activated by oxidative insults. Gene transfer experiments also demonstrated that expression of crmA in primary cultures of hippocampal neurons is protective against cell death induced by Abeta1-42. Together these results support the hypothesis that Abeta-induced apoptosis occurs through activation-induced programmed cell death.

the death of cultured hippocampal neurons (6), and A␤ is not toxic when immobilized as a neuronal substrate (9). These findings have led to the suggestion that aggregated A␤ might cross-link transmembrane plasma membrane receptors to initiate a death program, in a type of activation-induced programmed cell death (APCD) (10). Studies of A␤ effects on cells have documented changes in tyrosine phosphorylation of cellular substrates (11,12), suggesting that activation of signal transduction pathways might also be involved. Because A␤ causes oxidative stress in neurons, A␤ has been proposed to cause death through an oxidative mechanism (13), and antioxidants have sometimes (13) but not always (14) been reported to block A␤-induced cell death. Other studies have suggested that A␤ perturbs calcium homeostasis in neurons to cause cell death (15).
Recent studies have demonstrated that multiple pathways of apoptosis can be discerned by their differential sensitivity to blockade. In one model of APCD, Fas ligand binds to its receptor, causing receptor aggregation, recruitment of death domain-and death effector domain-containing proteins, and activation of a cascade of caspase proteases (16). APCD induced by Fas can be blocked by the viral serpin crmA (17), a caspase inhibitor whose physiological target is likely to be the apical caspase in the cascade (18), but it is insensitive to blockade by bcl-2 (Refs. 19 -21, but also see Refs. [22][23][24]. In contrast, apoptosis induced by the protein kinase inhibitor staurosporine (STS) is not inhibited by crmA (25), but is very efficiently blocked by bcl-2 (26). Other insults initiate apoptosis which is similarly sensitive to blockade by bcl-2 but insensitive to crmA, including etoposide, a DNA topoisomerase inhibitor (27).
In order to differentiate pathways used by apoptotic insults, stable transfectant PC12 cell lines expressing crmA or bcl-2 were generated. The results shown here demonstrate that APCD and non-APCD pathways of apoptosis differentially sensitive to inhibition by crmA and bcl-2 exist in PC12 cells. APCD induced by concanavalin A (ConA) is blocked by crmA, while apoptosis elicited by STS and oxidative insults is blocked by bcl-2. Importantly, A␤-induced cell death is blocked by crmA, suggesting that A␤ may cross-link cell surface receptors to engage an APCD apoptotic pathway.

EXPERIMENTAL PROCEDURES
Vectors-The complete coding sequence of human bcl-2 (provided by D. Hockenbery; Fred Hutchinson Cancer Research Center, Seattle, WA) (28) was subcloned into the pCDNA3 expression vector (Invitrogen) which uses the cytomegalovirus promoter to direct high levels of transgene expression in many types of cells. Similarly, a cDNA encoding the viral serpin crmA (provided by D. J. Pickup, Duke University) (29) was subcloned into pCDNA3. An IRES (internal ribosome entry site)-lacZ reporter sequence (30) was subcloned into the pHSVpuc amplicon (generously provided by F. Lim, Universidad Autonoma de Madrid). CrmA or bcl-2 cDNAs were subcloned upstream of the IRES-lacZ sequence to generate amplicons directing expression of bicistronic mRNAs.
PC12 Cell Toxicity Assays-Cells were passaged at 1e4 cells/cm 2 into 48-well tissue culture dishes coated with 50 g/ml poly-D-lysine and 20 g/ml type 1 collagen (Sigma). The following day, cells were rinsed and the medium was replaced with Opti-MEM (Life Technologies, Inc.) supplemented with 2 mM CaCl 2 . After 4 h, drugs were added, and 24 to 28 h later cell viability was assessed by nuclear morphology. Nuclei in live cells or cells previously fixed with 4% paraformaldehyde in phosphate-buffered saline were stained by the addition of 0.5 M SYTO11 (Molecular Probes), and cells were visualized with phase-contrast and epifluorescence microscopy. For each condition, live/dead cell counts were obtained from 2-3 fields of 3-4 wells. Cells with large nuclei containing uniformly stained chromatin were counted as live cells, while cells containing fragmented nuclei and/or condensed chromatin were counted as dead cells. None of the drugs used caused changes in the total cell number (live ϩ dead) over the course of the assay. Therefore, the number of live treated cells, expressed as the percentage of the number of live vehicle-treated cells, was used as a measurement of cell viability. Data shown are from representative experiments. Each experiment was repeated several times with similar results. In some experiments the ability of cells to oxidize 3-(4,5-dimethyl thiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) was used as a measurement of cell viability (32). Similar results were obtained regardless of the method used to assess cell viability when STS and ConA were used to induce cell death.
Western Blots-Cells were rinsed with Opti-MEM and total cell lysates were prepared by scraping cells into 1% SDS, 10 mM Tris, pH 7.5, 1 mM EDTA. Protein was estimated by the BCA assay (Pierce), and 1-15 g of total cellular protein was separated by 12% SDS-polyacrylamide gel electrophoresis and blotted onto Immobilon-P polyvinylidene difluoride (Invitrogen). Blots were blocked with 3% goat serum, 2% bovine serum albumin in TBS containing 0.1% Tween 20 and incubated with rabbit anti-bcl-2 (Santa Cruz 492 at 1:1000) in block. Antibody was detected using horseradish peroxidase-conjugated secondary antibody and ECL reagent (Amersham). Films were scanned and analyzed using NIH Image.
Primary Cultures of Hippocampal Neurons-Hippocampi of E18 rat embryos were dissected in CMF (calcium-and magnesium-free Hanks'balanced salt solution, containing 20 mM HEPES, 1 mM pyruvate, 4.2 mM sodium bicarbonate, and 0.3% bovine serum albumin), rinsed with CMF, and resuspended in a trypsin solution (0.125% trypsin in CMF containing 0.5 mM EDTA) for 8 min at 37°C. The trypsinization was stopped by the addition of DMEM containing 10% fetal calf serum, and the tissue was centrifuged at 200 ϫ g for 5 min. The resulting cell pellet was resuspended in 2 ml of culture medium (DMEM, Life Technologies, Inc., 12100-046, containing 20 mM HEPES, 26.2 mM sodium bicarbonate, 1 mM sodium pyruvate, and B27 supplement, Life Technologies, Inc.). Following trituration through fire-polished Pasteur pipettes with the diameter maximally 50% constricted, the cell suspension was filtered through a 40-m cell strainer (Falcon), and viable cells were counted using trypan blue. Cells were plated at 5-8e4 cells/cm 2 in 48-well tissue culture dishes (Costar).
Transduction of Neurons with Herpes Simplex Virus-1 Amplicons and Toxicity Assays-Amplicons were packaged using replication incompetent 5dl1.2 helper virus (obtained from P. A. Schaffer, University of Pennsylvania) and 2-2 cells (provided by R. Sandri-Goldin, University of California, Irvine) and purified through sucrose gradient centrif-ugation as described previously (33). Viral vectors were titered on PC12 cells as described (33). Hippocampal neurons were infected at a multiplicity of infection of 0.1-0.01 with crmA-IRES-lacZ, bcl2-IRES-lacZ, or IRES-lacZ vectors after 2 days in culture. Cells were treated with drugs the next day and fixed 24 h later by underlay with 4% paraformaldehyde in phosphate-buffered saline containing 5% sucrose for 30 min. Fixed cultures were blocked in TBS containing 0.3% Tween 30, 3% goat serum, and 2% bovine serum albumin and then incubated with 40-1a (1:4000 ascites) anti-␤-galactosidase (developed by Joshua Sanes and obtained from the Developmental Studies Hybridoma Bank maintained by the University of Iowa Department of Biological Sciences under contract NO1-HD-7-3263 from the National Institute of Child Health and Human Development, National Institutes of Health). ␤-Galactosidase immunoreactivity was detected by incubation with 7 g/ml cy2antimouse IgG (Jackson Immunoresearch Laboratories). Nuclear morphology was assayed by inclusion of 1.25 g/ml Hoechst 33258 (Sigma) in the secondary antibody incubation. Cells were visualized by epifluorescence microscopy. Noninfected cells and cells infected with 5dl1.2 helper virus were uniformly negative for ␤-galactosidase immunoreactivity. Cells transduced by amplicons, identified by the presence of ␤-galactosidase immunoreactivity, were counted and their nuclear morphology (normal or apoptotic) recorded. Results are presented as the number of infected cells with normal nuclei after drug treatment as a percentage of the number of infected cells with normal nuclei after vehicle control treatment.

RESULTS
In preliminary experiments to establish dose-response curves for cell death, PC12 cells were exposed to increasing concentrations of STS, ConA, A␤ 1-42 , A␤ [25][26][27][28][29][30][31][32][33][34][35] , or hydrogen peroxide in Opti-MEM, a low protein-containing chemically defined medium. Cell viability was assessed after 24 h by nuclear morphology visualized with the fluorescent dye SYTO11, since this method allows clear visualization of apoptotic nuclei (Fig. 1). Over 95% of control cells were viable after serum withdrawal into Opti-MEM, but exposure to these insults caused dose-dependent apoptotic death of the cells, evidenced by somal shrinkage, plasma membrane blebbing, chromatin condensation, and nuclear fragmentation. Drug concentrations that caused 50 -90% of the cells to die over the course of 24 h were chosen for use in subsequent assays. Preincubation with the cell permeable caspase inhibitor zVAD-fmk completely blocked cell death induced by these concentrations of ConA, STS, A␤  , and A␤ [25][26][27][28][29][30][31][32][33][34][35] , and partially blocked cell death induced by this concentration of hydrogen peroxide, consistent with the involvement of apoptotic pathways in the observed cell death (Fig. 2).
To differentiate pathways of apoptosis in PC12 cells, cells were transfected with an expression vector encoding bcl-2 or an empty vector control, and stable transfectants that expressed various levels of bcl-2 were selected (Fig. 3A). These cells lines were tested for their susceptibility to 2 prototypical apoptotic insults: STS and ConA. Cell lines that produced moderate levels of bcl-2, bcl-2#11 and bcl-2#12, were better protected from STS-induced apoptosis than from ConA-induced apoptosis (Fig. 3B). Cell lines expressing approximately 6-fold higher levels of bcl-2 (bcl-2#3 and bcl-2#10), however, were protected from apoptosis induced by both STS and ConA.
An expression vector encoding crmA was used to transfect PC12 cells, and stable transfectants were selected and screened for their susceptibility to apoptotic insults. All of the crmAtransfected cell lines were protected from ConA-induced apoptosis (not shown). Two lines, crmA#3 and crmA#5, that showed the highest level of protection from ConA were used in subsequent assays. Neither crmA#3 nor crmA#5 cells were protected from STS-induced apoptosis (Fig. 4). Together, these results suggested that at least 2 pathways of apoptosis exist in PC12 cells. One pathway, activated by STS, is preferentially blocked by bcl-2 and is not blocked by crmA. The second pathway, activated by ConA, is blocked by crmA but is less efficiently blocked by bcl-2.
Cell death induced by A␤ has been associated with oxidative damage to cells (35)(36)(37), yet under at least some assay conditions antioxidants protect against hydrogen peroxide-and ironinduced cell death, but do not protect against A␤-induced cell death (14). This suggests that the mechanism of A␤ toxicity is less dependent on oxidative damage. To determine whether bcl-2 and crmA transfectants were protected from cell death induced by oxidative stress, cell survival was assessed after exposure to hydrogen peroxide (Fig. 6A). As expected, cells expressing bcl-2 were protected from cell death induced by this oxidative insult, but crmA transfectants were not protected from cell death. Moreover, in other experiments cell lines expressing moderate levels of bcl-2 were observed to be protected from cell death induced by the oxidative insults peroxynitrite and FeSO 4 , while crmA transfectants were not protected from death. Cell viability after exposure to 100 M peroxynitrite was 95.4 Ϯ 2.8% in bcl-2 transfectants, 42.1 Ϯ 1.5% in crmA trans- fectants, and 47.6 Ϯ 0.8% in control transfectants (mean Ϯ S.D., n ϭ 2, p Ͻ 0.01 survival of bcl-2 transfectants versus control transfectants), while cell viability after exposure to 120 M FeSO 4 was 94.5 Ϯ 3.9% in bcl-2 transfectants, 46.3 Ϯ 3.3% in crmA transfectants, and 49.5 Ϯ 4.8% in control transfectants (mean Ϯ S.D., n ϭ 2, p Ͻ 0.01 survival of bcl-2 transfectants versus control transfectants). These data show that bcl-2 levels sufficient to block apoptotic pathway(s) activated by oxidative insults do not block the apoptotic pathway activated by A␤ 1-42 , while crmA does block the apoptotic pathway activated by A␤ 1-42 , but does not block apoptotic pathway(s) activated by oxidative insults. Therefore, these results strongly suggest that oxidative stress does not mediate the initial activation of an apoptotic program by A␤  . Accordingly, antioxidants including GSH (Fig. 6B) and propyl gallate (data not shown) did not block cell death induced by A␤ 1-42 in these cells.
Taken together, these results indicate that multiple apoptotic pathways can be discerned in PC12 cells, and that the apoptotic pathway activated by A␤ 1-42 is susceptible to inhibition by crmA. However, although PC12 cells are often used to model aspects of neuronal physiology, there is always the possibility that cell death might be differentially regulated in immortalized "neuronal" cell lines and neurons. Therefore, in order to confirm that crmA blocks the apoptotic pathway activated by A␤  in neurons, primary cultures of hippocampal neurons were transduced with crmA using a herpes simplex virus amplicon containing an IRES-lacZ reporter. Infected cells were identified by their expression of ␤-galactosidase and their viability determined by nuclear morphology (Fig. 7). Cells infected with the control IRES-lacZ amplicon and uninfected cells were similarly susceptible to apoptosis induced by ConA, STS, A␤ 1-42 , and A␤ [25][26][27][28][29][30][31][32][33][34][35] (Fig. 8). However, as predicted, neurons transduced with crmA were protected from cell death induced by ConA and A␤ 1-42 , but were not protected from cell death induced by STS or A␤ [25][26][27][28][29][30][31][32][33][34][35] (Fig. 8). In other experiments, hippocampal neurons transduced with bcl-2 were protected from cell death induced by each of these insults (data not shown). Because the IE 4/5 promoter used in the amplicon directs a high level of gene expression in neurons (38), these results appear consistent with the results obtained in PC12 cells expressing high levels of bcl-2.

DISCUSSION
The accumulation of plaques containing A␤ is an invariant feature of AD pathology, and there is abundant evidence suggesting that A␤ contributes to the etiology of AD. Most significantly, chromosome 21-linked familial AD is caused by amyloid precursor protein alleles with mutations in or near the A␤ coding sequence (39), and transgenic mice expressing these amyloid precursor protein alleles exhibit pathological changes and memory deficits reminiscent of AD (40,41). Furthermore, AD is characterized by a profound loss of neurons in susceptible regions of the brain (1), and A␤ causes the apoptotic death of cultured neurons (5,7,8). Therefore, understanding the mechanisms by which A␤ induces apoptosis may be relevant to discovery of clinical interventions to delay or alleviate AD.
We have previously suggested that cell death induced by A␤ may represent a type of APCD (10,42). Neurons have been shown to be susceptible to APCD: transforming growth factor ␤ causes the apoptotic death of cerebellar granule neurons (43), and ConA, a lectin which binds mannose and cross-links glycoproteins on the cell surface, causes the apoptotic death of cortical neurons (42). Other insults including uv radiation (44) and growth factor withdrawal (45,46) may also cause apoptosis through APCD. However, apoptosis has also been demonstrated in neurons exposed to oxidative stress (47), etoposide (48), heat shock (49), and decreased extracellular Kϩ (50), indicating that non-APCD apoptotic pathways also exist in neurons. Alternate pathways of apoptosis can be discerned by their differential sensitivity to blockade by bcl-2, crmA, and various caspase inhibitors (20,51). Anti-apoptotic members of the bcl-2 family of proteins are very effective in blocking apoptosis induced by oxidative stress (30,52) and STS (26), while the viral serpin and caspase inhibitor crmA effectively inhibits Fas-and TNF-mediated apoptosis (17). Both bcl-2 and crmA, however, block apoptosis prior to activation of caspase-3 (25,53). Therefore, although multiple pathways are involved in the induction of apoptosis, these pathways converge on the activation of caspases which are the effectors of apoptosis.
The results presented here demonstrate that in PC12 cells, ConA and STS activate separate apoptotic pathways that are differentially blocked by crmA and bcl-2. Because the physiological target of crmA is likely to be caspase-8, the apical caspase activated by Fas and TNFR (18), the ability of crmA to inhibit ConA-induced apoptosis in PC12 cells suggests the possibility that ConA might cross-link death domain containing receptors of the Fas/TNFR family to initiate a death program. In addition to inhibiting Fas-and TNF-initiated apoptosis, crmA inhibits anoikis (54), apoptosis induced by NGF withdrawal from sympathetic neurons (55), and apoptosis induced by serum withdrawal from PC12 cells. 2 These insults may also cause death through inappropriate activation of cell surface receptors. For example, death produced by NGF withdrawal has been proposed to be mediated by the low affinity NGF receptor (p75 NGFR), a member of the fas/TNF superfamily of receptors that has been reported to cause apoptosis in the absence of NGF (45).
In a previous study, overexpression of bcl-2 did not inhibit A␤ 25-35 -induced death of PC12 cells or of human neuroblastoma IMR-5 cells (58). Cell death was assessed by the MTT assay, however, which has been shown to be an inaccurate indicator of cell death induced by A␤ (59,60). The results shown here demonstrate that the ability of bcl-2 to inhibit A␤-induced death of PC12 cells is dependent on its level of expression: relatively high levels of bcl-2 expression are required to block apoptosis induced by A␤ [25][26][27][28][29][30][31][32][33][34][35] or A␤  . Interestingly, moderate levels of bcl-2 sufficient to inhibit apoptosis induced by oxidative insults in PC12 cells were not able to inhibit apoptosis induced by A␤  . Moreover, and in agreement with the results we have obtained in primary cultures of hippocampal neurons (14), antioxidants did not block apoptosis induced by A␤  or A␤ [25][26][27][28][29][30][31][32][33][34][35] . In contrast, a recent study demonstrated that A␤ 25-35 toxicity in PC12 cells was blocked by antioxidants (37). As previously discussed (14), discrepancies in the effectiveness of antioxidants in blocking A␤ toxicity may be related to methodological differences between experiments.
In the experiments described here, cells were exposed to insults in Opti-MEM, a medium which supports the growth of the cells for at least several days. These cells might be healthier and more resistant to oxidative damage than cells assayed after serum and NGF withdrawal into RPMI (37), since PC12 cells deprived of trophic support in this manner undergo apoptosis within 2 to 3 days (61,62).
There has been some ambiguity in the literature as to whether bcl-2 and related anti-apoptotic members of this fam-2 K. J. Ivins, unpublished observations. ily of proteins are able to block Fas-or TNF-induced apoptosis (19 -24). The results here demonstrate that PC12 cells expressing moderate levels of bcl-2 were preferentially protected against apoptosis induced by oxidative insults and STS, but cells expressing higher levels of bcl-2 were also completely protected against apoptosis induced by ConA. These results suggest that the level of bcl-2 expression may be critical to its ability to suppress APCD, and that the ability of bcl-2 to block Fas-or TNF-induced apoptosis is likely to require high levels of expression.
It is not known whether apoptotic pathways in neurons are identical to those demonstrated here in PC12 cells. However, because apoptotic execution may involve inappropriate activity of proteins involved in the regulation of the cell cycle (63,64), it has been suggested that differentiated and cycling cells use similar apoptotic programs, and what varies is whether new protein synthesis is required or whether the necessary proteins are available because they are constitutively made in cycling cells (63). In fact, our results demonstrate that crmA protects hippocampal neurons as well as PC12 cells from death induced by ConA and A␤ 1-42 , but does not block cell death induced by STS or A␤ [25][26][27][28][29][30][31][32][33][34][35] in either type of cell culture. Therefore, our results suggest that central nervous system neurons and PC12 cells use similar apoptotic pathways, and that A␤ 1-42 causes APCD in neurons as well as in PC12 cells.