Alix, a Protein Regulating Endosomal Trafficking, Is Involved in Neuronal Death*

Alix/AIP1 is a cytoplasmic protein, which was first characterized as an interactor of ALG-2, a calcium-binding protein necessary for cell death. Alix has also recently been defined as a regulator of the endo-lysosomal system. Here we have used post-mitotic cerebellar neurons to test Alix function in caspase-dependent and -independent cell death. Indeed, these neurons survived when cultured in 25 mm potassium-containing medium but underwent apoptosis soon after the extracellular potassium was lowered to 5 mm. In agreement with other studies, we show that caspases are activated after K+ deprivation, but that inhibition of these proteases, using the pancaspase inhibitor boc-aspartyl(OMe)-fluoromethylketone, has no effect on cell survival. Transfection experiments demonstrated that Alix overexpression is sufficient to induce caspase activation, whereas overexpression of its C-terminal half, Alix-CT, blocks caspase activation and cell death after K+ deprivation. We also define a 12-amino acid PXY repeat of the C-terminal proline-rich domain necessary for binding ALG-2. Deletion of this domain in Alix or in Alix-CT abolished the effects of the overexpressed proteins on neuronal survival, demonstrating that the ALG-2-binding region is crucial for the death-modulating function of Alix. Overall, these findings define the Alix/ALG-2 complex as a regulator of cell death controlling both caspase-dependent and -independent pathways. They also suggest a molecular link between the endo-lysosomal system and the effectors of the cell death machinery.

A bridge between the endo-lysosomal system and cell death has been suggested by numerous observations dealing with apoptosis or with the so-called type II autophagic cell death, which is independent of caspases (1)(2)(3)(4)(5). The endo-lysosomal system is composed of a series of intracellular compartments within which endocytosed molecules and redundant cellular material are hydrolyzed (6). Endocytosed material tends to flow vectorially through the system from early endosomes, along late endosomes into lysosomes, whereas phagocytosis and autophagy provide alternative entry points.
Multivesicular bodies represent a subset of late endosomes characterized by their multiple intralumenal vesicles, which bud from the limiting membranes together with ubiquitinated transmembrane proteins destined for degradation inside lysosomes. In Saccharomyces cerevisiae, ubiquitinated membrane proteins en route to the vacuole are recognized and sorted in the multivesicular bodies by three hetero-oligomeric complexes (ESCRT-I, ESCRT-II, and ESCRT-III) composed of Vps 1 proteins (7)(8)(9). Recently, several publications have characterized mammalian homologues of Vps proteins and described their mutual interactions. They show that ESCRT-I and -III are bridged by a 97-kDa cytoplasmic protein named Alix/AIP1 (10 -13). They also demonstrate that the GAG protein from membrane-containing viruses, such as HIV, binds to Alix/AIP1, thereby recruiting the ESCRT machinery to allow budding of the virus from the cell surface (11)(12)(13). Alix has previously been shown to be highly enriched in phagosomes (14) and exosomes, the latter being extracellular vesicles secreted upon fusion of the limiting membrane of the multivesicular bodies with the plasma membrane (15). We have shown that the 150-amino acid long proline-rich domain of the C-terminal part of Alix (Alix-CT) interacts with the SH3 domain of endophilins (16), which are lysophosphatidic acid acyltransferases capable of modifying the lipid bilayer curvature during endocytosis (17). Moreover, the Alix proline-rich domain interacts with CIN85/ SETA/Ruk (18), which regulates endocytosis of ubiquitinated tyrosine kinase receptors (19,20). These different observations suggest that Alix regulates the endo-lysosomal pathway from the cell surface down to late endosomes.
Several other observations suggest a role for Alix in controlling cell death. The protein was first characterized as an interactor of the penta-EF hand calcium-binding protein ALG-2 (apoptosis-linked gene 2) (21,22). ALG-2 seems necessary for death to occur as inhibition of its expression in a T cell line blocked apoptosis (23) without impeding caspase-3 activation (24). However, death was not impaired in T cells from ALG-2deficient mice, suggesting that other proteins of similar function may replace ALG-2 (25). In addition, another binding partner of Alix, Cin85, not only regulates endocytosis but also * This work was supported in part by INSERM, the University Joseph Fourier, grants from the Association pour la Recherche contre le Cancer and the Association Française contre les Myopathies. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  influences cell death of astrocytes and neurons (18,26). A final argument for a possible role of Alix in cell death comes from our recent demonstration of its up-regulation in neurons degenerating after kainate-induced epileptic seizures (53).
Here we have used potassium depletion-induced death of cerebellar granule neurons (CGN) as a paradigm of neuronal death to test a role of Alix in this phenomenon. We demonstrate that both caspase-dependent and -independent mechanisms underlie death of these neurons because inhibition of caspases does not impair death induced by potassium deprivation. We show that Alix overexpression is sufficient to induce caspase activation and nuclear alteration leading to death of neurons cultured in high potassium. In turn, Alix-CT inhibits caspase activation after potassium deprivation; it also inhibits the caspase-independent program because transfected neurons survive in low K ϩ . Finally, we identify the region of Alix necessary for binding ALG-2 and demonstrate that this site, but not the one binding endophilins, is required for Alix to control cell death. The complex Alix/ALG-2 may thus represent a critical link between the endo-lysosomal pathway and the cell death effectors.

EXPERIMENTAL PROCEDURES
Cell Culture Media-All culture media were based on Dulbecco's modified Eagle's medium (DMEM, Invitrogen) containing 10 unit/ml penicillin, 10 g/ml streptomycin, 2 mM L-glutamine, and 10 mM HEPES (K5 medium). KCl was added to a final concentration of 25 mM (K25 medium) and supplemented with 10% fetal bovine serum (K25ϩS medium).
Neuronal Culture-Primary cultures of CGN were prepared from 6-day-old S/IOPS NMRI mice (Charles River Laboratories), as described previously (27,28), with some modifications. The cerebella were removed, cleared of their meninges, and cut into 1-mm pieces. They were then incubated at 37°C for 10 min in 0.25% trypsin-EDTA (Invitrogen) in DMEM. Trypsin was inactivated with an equal volume of K25ϩS medium and 3000 units/ml DNase I (Sigma) before dissociation by triturating by using flame-polished Pasteur pipettes in DMEM containing trypsin inhibitor and 300 units/ml DNase I. Dissociated cells were centrifuged for 5 min at 500 ϫ g. The pellet was resuspended in fresh K25ϩS medium and cells were plated onto poly-D-lysine (10 g/ml, Sigma) precoated 60-mm dishes or 24-well plates at a density of 3 ϫ 10 5 cells/cm 2 . The cerebellar granule neurons were grown in K25ϩS medium in a humidified incubator with 5% CO 2 /95% air at 37°C. Cytosine-␤-D-arabinoside (10 M, Sigma) was added after 1 day in vitro (DIV) to prevent the growth of non-neuronal cells. After 4 DIV, K25ϩS medium was replaced with serum-free K25 medium to eliminate the 20% of cells dying within a few hours after serum removal (28). Twentyfour hours later, cells were changed to K5 medium or maintained in K25 medium. In some experiments, 10 M BAF (boc-aspartyl(OMe)-fluoromethylketone; Bio-Rad), a general caspase inhibitor, was added directly to the medium.
Determination of Cell Viability-Survival was estimated 0, 3, 6, 12, and 24 h after switching the cultures from K25 medium to K5 medium using the Alamar blue assay (BIOSOURCE) according to the instructions of the manufacturer. In other cases, cells were fixed in 4% paraformaldehyde in PBS for 20 min at 4°C and stained with Hoechst 33342 (4 g/ml, Sigma) in PBS for 30 min at 37°C. Cell viability was then scored on the basis of nuclear morphology; condensed or fragmented nuclei were taken to indicate cell death.
Plasmids-All constructions used are based on the pCi vector (Promega) and have been described by Chatellard-Causse et al. (16). All cDNAs had a tag epitope and were coded for the following proteins: Alix, ALG-2, Alix ⌬PP14 (deleted from amino acids 748 -761), Alix ⌬PGY (deleted from amino acids 802-813), Alix-NT (N-terminal half of Alix ending at residue 434), Alix-CT (corresponds to the first nine amino acids of Alix linked to the C-terminal half of Alix starting at amino acid 468), and Alix-CT ⌬PGY (Alix-CT deprived of the PGY domain), and GFP (green fluorescent protein) cDNA as control for the toxicity of transfection.
Transfection of CGN-Transient transfection of neurons in primary culture was performed by the calcium phosphate coprecipitation method developed by Xia et al. (29) and modified by Finkbeiner and Greenberg (30). Transfection was performed on neurons cultured 4 days on 14-mm glass coverslips. K25ϩS medium was removed, and the cells were incubated for 1-2 h with fresh DMEM supplemented with 10 mM MgCl 2 in 5 mM HEPES, pH 7.4. Meanwhile, the DNA/calcium phosphate precipitate was prepared and allowed to form for 25-30 min at room temperature before addition to the cultures. Eight micrograms of total plasmid DNA were used for each coverslip. After 1 h of incubation with the DNA/calcium phosphate precipitate, the cells were "shocked" for 1 min with 2ϫ HeBS buffer (274 mM NaCl, 10 mM KCl, 1.4 mM Na 2 HPO 4 , 7H 2 O, 15 mM D-glucose, 42 mM HEPES, pH 7.6), 10 mM MgCl 2 in 5 mM HEPES, pH 7.4, and 2% Me 2 SO. The cells were then washed three times with DMEM, and K25 medium was added to each coverslip. They were then returned to the 5% CO 2 incubator at 37°C and cultured for the indicated time. The transfection efficiency was determined after immunolabeling with an anti-tag antibody; it was typically between 15 and 20%.
Immunocytochemistry-Neurons were washed twice with PBS and fixed in 4% paraformaldehyde in PBS for 20 min at 4°C. Cells were permeabilized with 0.02% saponin in Tris-buffered saline for 30 min and then incubated with antibodies as follows: mouse anti-Myc (Santa Cruz Biotechnology, 1:100), mouse and rabbit anti-FLAG (Sigma, 1:100), rabbit anti-phosphoserine 63 c-Jun and rabbit anti-phospho p38 MAPK (New England Biolabs, 1:100). Primary antibodies were revealed with either goat anti-mouse or anti-rabbit immunoglobulin G coupled to fluorochrome Alexa 594 or Alexa 488 (Molecular Probes) diluted 1:1000 in saponin-Tris-buffered saline. The cultures were mounted in Mowiol (Calbiochem), and the fluorescence was observed under an Axiovert microscope (Zeiss) connected to a CDD camera or a confocal laser scanning microscope (LSM410, Zeiss). Image analysis was performed by using Metamorph TM software (Universal Imaging Corp.).
Detection of Active Caspases in Living Cells-CaspaTag TM fluorescein caspase (VAD) activity kit (Intergen) allows the detection of active caspases (caspase-1-9) in living cells through the use of a carboxyfluorescein-labeled fluoromethylketone pan-caspase inhibitor (fam-VADfmk). This probe enters the cell and binds irreversibly to activated caspases. After 24 h in either K25 or K5 medium, CGN were incubated for 1 h at 37°C with fam-VAD-fmk in accordance with the supplier's instructions and visualized under an epifluorescence microscope.
Overlay Assays-HEK293 cells were prepared as described in Chatellard-Causse et al. (16). Proteins (10 g/lane) were run in 8% SDS-PAGE and blotted to nylon membranes. Membranes were saturated in blocking solution (Tris-buffered saline, 0.1% Tween, 5% nonfat dry milk, and 2 mM CaCl 2 ) for 1 h at room temperature and incubated with Myc-ALG-2 recombinant protein in blocking solution. After 2.5 h, membranes were washed in TBS, 0.1% Tween 20, and 2 mM CaCl 2. The bound recombinant protein was revealed with an anti-Myc monoclonal antibody followed by a goat anti-mouse horseradish peroxidase-conjugated secondary antibody; then the SuperSignal detection method was employed (Pierce).

Potassium Depletion-induced Cell Death of CGN-CGN
can be grown in the absence of serum in a medium containing 25 mM KCl (K25 medium) but undergo apoptosis soon after changing to 5 mM KCl (K5 medium) (27). Cell viability, estimated by the Alamar blue assay (Fig. 1A) or by scoring pyknotic nuclei (not shown), was reduced by 50% 24 h after potassium depletion. In agreement with previous studies (31-33), addition of the general caspase inhibitor BAF (10 M) blocked nuclear fragmentation but had no effect on nuclear condensation (not shown) and cell survival (Fig. 1A).
Potassium depletion was followed rapidly by activation of caspases. Caspase-3, a central caspase executioner, was activated within 4 h, as shown by the gradual increase in the level of the active p17 form (Fig. 1B). Determination of caspase-3 proteolytic activity in cell lysates demonstrated that 6 h after K ϩ lowering, the enzymatic activity increased ϳ20 times above the control level (Fig. 1C). We also observed that PARP, a specific caspase-3 substrate, was cleaved into a characteristic 89-kDa fragment (Fig. 1D). Caspase-3 activity was blocked when BAF was added to the culture medium at the time of K ϩ depletion; indeed, the activity in cell lysates from neurons cultured in low K ϩ medium supplemented with BAF was almost identical to the activity in lysates from cells grown in K25 medium (Fig. 1C). Furthermore, besides caspase-3, K ϩ withdrawal also triggered activation of caspase-2, -6, -8, and -9, which were all blocked by BAF (data not shown). These observations, showing that caspase inhibition is not sufficient to allow survival, demonstrate that both caspase-dependent and -independent programs are at work during the death of CGN triggered by potassium depletion.
Alix Overexpression Is Sufficient to Induce Apoptosis of CGN-To evaluate a possible involvement of Alix in neuronal death, CGN were transfected with Alix and left in K25 medium. Ten hours after transfection, overexpressed Alix was homogeneously distributed throughout the cell body and neurites ( Fig. 2A). Determination of cell viability demonstrated that the transfected neurons underwent cell death; 24 h after transfection, 50% of the Alix-overexpressing neurons grown in 25 mM KCl were dead or dying, whereas neurons transfected with GFP (control vector) were unaffected (Fig. 2B). At this time, caspase activity could be detected in situ using CaspaTag TM , and nuclear condensation was clearly visible in Alix-overexpressing neurons (Fig. 2C-E). Addition to the culture medium of the general caspase inhibitor BAF blocked CGN death induced by overexpressed Alix (Fig. 2B). Thus, overexpression of Alix is sufficient to induce caspase-dependent neuronal death in CGN cultured in the presence of survival signals. C-Jun N-terminal kinases (JNK) and p38 MAPK are important regulators of the cell death program in post-mitotic neurons. In particular, the JNK pathway and one of its major downstream targets, the transcription factor c-Jun, are thought to play an essential role in the death of CGN (34 -39). We performed immunofluorescence analysis of c-Jun and p38 MAPK activation to test whether Alix overexpression launches apoptosis by activating these cascades (Fig. 3A-D). Although phospho-c-Jun and -p38 MAPK immunoreactivity were clearly detected in CGN grown in K5 medium (Fig. 3E) or treated with anisomycin, a known p38 MAPK activator (Fig. 3F), there was no evidence to suggest that JNK and p38 MAPK were activated in Alix-overexpressing neurons. Hence, Alix does not induce death of CGN by activating either JNK-or p38 MAPK-signaling pathways.
Overexpression of Alix-CT Protects CGN from Death Induced by Potassium Withdrawal-Overexpression of the C-terminal half of Alix led to the formation of numerous cytoplasmic vacuoles with a diameter ranging from 0.5 to 2 m (Fig. 4, A and C and Fig.5). These vacuoles are reminiscent of those seen in HEK293 cells overexpressing Alix-CT together with endophilins (16). Most Alix-CT-expressing neurons had a healthy nuclear appearance after K ϩ lowering, unlike nearby untransfected neurons (Fig. 4E). Twenty-four hours after potassium deprivation, neurons with a condensed nucleus represented only 15% of Alix-CT-expressing cells, in contrast to ϳ50% for untransfected, Alix, or GFP-transfected cells (Fig. 4B). The labeling with CaspaTag TM showed no detectable caspase activity in most Alix-CT neurons, in contrast to nontransfected neurons (Fig. 4D). Overexpression of Alix-CT thus protects CGN from death induced by potassium withdrawal.
Interestingly, in the few Alix-CT-overexpressing neurons showing detectable caspase activity, the CaspaTag TM staining was concentrated in spots colocalized with Alix-CT (Fig. 5A). Confocal analysis of these cells demonstrated that the CaspaTag TM fluorescence was trapped in the vacuoles delineated by Alix-CT (Fig. 5B), suggesting that active caspases were concentrated inside. We made similar observations using another CaspaTag TM (DEVD) specific for caspase-3 (not shown), demonstrating that caspase-3 is one of the active caspases found inside the vacuoles. It is noteworthy that these neurons always showed nuclear alterations typical of cells undergoing apoptosis.
ALG-2 Binds to a PGY Repeat from Alix Proline-rich Domain-We used a far Western blotting approach to determine the binding site of ALG-2 on Alix. For this, lysates from HEK293 cells overexpressing different mutants of Alix were run on SDS-PAGE and blotted onto nylon membranes. Purified Myc-tagged recombinant ALG-2 was then incubated on the membrane, and its binding was revealed with an anti-Myc antibody (Fig. 6A). The recombinant ALG-2 bound to overexpressed full-length Alix, to its C-terminal half, and to Alix lacking the site of interaction with endophilins (Alix ⌬PP14), but not to Alix-NT or to Alix ⌬PGY, which lacks residues 802-813 (Fig. 6A). The use of an anti-FLAG antibody showed that Alix ⌬PGY, Alix ⌬PP14, and Alix were all expressed at the same level (Fig. 6B). These results demonstrate that Alix amino acids 802-813 are necessary for ALG-2 binding.
Deletion of the ALG-2-binding Site Abrogates Alix Ability to Modulate CGN Survival-In cultures kept in K25 medium, neurons transfected with Alix ⌬PP14, lacking the endophilinbinding site, underwent apoptosis in the same way as Alixoverexpressing neurons, because about 50% died in less than 24 h (Fig. 7A). In contrast, the survival of CGN expressing large amounts of Alix ⌬PGY was not significantly different from control GFP-overexpressing neurons (Fig. 7A). These results indicate that apoptosis induced by overexpressed Alix is strictly dependent on the integrity of its ALG-2 interacting site. As shown in Fig. 7A, overexpression of ALG-2, in contrast to that of Alix, had no deleterious effects on neuronal survival.
We then addressed the question of whether the capacity of Alix-CT to bind ALG-2 is required for Alix-CT to block CGN death. CGN were transfected with an expression vector encoding Alix-CT lacking the ALG-2-binding domain (Alix-CT ⌬PGY) and deprived of K ϩ 24 h later. As illustrated in Fig. 7B, no significant difference in cell survival was recorded between control GFP-expressing neurons and Alix-CT ⌬PGY-expressing neurons: about 50% of the latter were dead or dying 24 h after the switch to K5 medium. Thus, these results demonstrate that the neuroprotective effect of Alix-CT on CGN apoptosis requires the region binding ALG-2.

DISCUSSION
Alix is a protein recently shown to regulate the endo-lysosomal system (10,16). This protein is also suspected to regulate neuronal death; we recently demonstrated that it is up-regulated in degenerating rat hippocampal neurons after epileptic seizures induced by kainate treatment (53). Here we show that enforced expression of Alix is sufficient to induce caspase activation and, thereby, death of cultured cerebellar granule neurons. This death, which displays all morphological features of apoptosis, is truly caspase dependent as, unlike that seen upon potassium depletion, it can be effectively blocked by the addition of a pan-caspase inhibitor to the culture medium. Neither of the stress-activated kinases tested (JNK and p38 MAPK) were activated, demonstrating that the pro-apoptotic function of Alix is not simply due to neuronal stress. Alix overexpression has recently been reported to induce detachment of cells from the substrate, and it even enhanced detachment-induced death (anoïkis) of HeLa cells (40). In our hands, Alix had no effect on neuron attachment to the substrate. Also noteworthy is the fact that the primary neurons that we transfected are post-mitotic, ruling out that the pro-apoptotic effect of Alix may be caused by its previously demonstrated capacity to induce G 1 arrest (41).
In vivo, a simple rise in intracellular Alix concentration is not sufficient to cause neuronal death, because, during avian development, the protein expression starts in motoneurons during axon outgrowth, well before the onset of programmed cell death (42). We hypothesized that ALG-2, a calcium-binding protein necessary for cell death (23) which binds in a calciumdependent way to the C-terminal proline-rich region of Alix (21,22), could regulate its pro-apoptotic activity. To test this, we first characterized the ALG-2 binding site by making Alix deletion mutants and evaluating their capacity to bind ALG-2 in a far Western. We found that ALG-2 could not bind to Alix ⌬PGY, which lacks amino acids 802-813 (PPYPTYPGYP-GYPGY), whereas it was still able to bind to Alix lacking other discrete regions of the proline-rich domain. We chose this region based on its resemblance with the PGY-biased amino acid composition of the N-terminal parts of annexins VII and XI, which also bind ALG-2 (43,44). In fact, the N-terminal region of annexin XI contains a PSYPGYPG sequence that is almost identical to the sequence found in Alix. The ALG-2 binding region is distinct from the endophilin-binding site that we previously characterized (amino acids 748 -761, deleted in Alix ⌬PP14; Ref. 16). It is also distinct from the PTPAPR sequence (amino acids 740 -745), which fulfills the PxPxPR consensus, recently described as necessary for high affinity binding to the SH3 domain of CIN85/SETA/Ruk (45) and from the PSAP (amino acids 717-720) required for Tsg101 binding (13). In overexpression experiments, Alix ⌬PGY, unlike Alix ⌬PP14, was unable to induce caspase activation, strongly suggesting that binding of ALG-2, but not binding of endophilins, is required for Alix pro-apoptotic activity. Because the binding of ALG-2 to Alix is strictly dependent on Ca 2ϩ , one may hypothesize that ALG-2 translates calcium fluctuations, induced by K ϩ deprivation, through Alix into a cell death response. The link between calcium and cell death is a matter of numerous publications (for review, see Refs. 46 and 47), and it is now clear that even non-disruptive changes in Ca 2ϩ homeostasis and calcium compartmentalization may regulate cell survival.
The Alix-ALG-2 couple may be another effector of Ca 2ϩ signaling in cell death.
It is difficult to predict a mechanism of action for Alix. It is noteworthy, however, that besides ALG-2, the proline-rich domain region interacts with CIN85, which has already been implicated in apoptosis (18,26), as its overexpression induced death of sympathetic neurons (26). In this latter case, it seems that CIN85 binds to and inhibits the type IA PI3 kinase, which is required for nerve growth factor survival activity (26). In agreement with several publications (48,49), we have observed that a potent PI3 kinase inhibitor, LY294002, induces apoptosis of cerebellar neurons kept in high K ϩ (not shown). In the near future, we will examine whether Alix deprived of its potential CIN85-binding site is still able to induce neuronal death, thereby testing the possibility that the Alix/ALG-2 complex recruits CIN85 to drive cell death.
Overexpression of Alix-CT blocked caspase activation in the majority of transfected neurons deprived of K ϩ . Alix-CT could act as a dominant-negative mutant by binding to endogenous Alix (22), which is known to interact with itself through its C-terminal half (21). Alix-CT could also directly affect endosomal function, because it possesses the site of binding to ESCRT-I but lacks the ESCRT-III interacting site (amino acids 1-176 of Alix). We have shown that overexpressing Alix-CT induces cytoplasmic vacuolization, and several Alix N-terminal deletion mutants lacking the ESCRT III-binding region were recently found to block budding of viruses at the cell surface (11)(12)(13). This latter observation was thought to reflect the impairment by truncated Alix of the cell ESCRT machinery normally recruited by viruses to bud at the cell membrane. Overexpressing Alix-CT also led to the formation of vacuoles in neurons. In potassium-depleted cultures, active caspases filled the entire soma, whereas in the few Alix-CT-expressing neurons in which we detected active caspases, these were concentrated inside the vacuoles. The vacuoles could represent a way for neurons to inactivate the proteases, which would explain why caspase activity was absent in the vast majority of Alix-CT-FIG. 6. The PGY domain of Alix is necessary for binding to ALG-2. A, 10 g of proteins from lysates of HEK293 cells transfected with FLAG-tagged Alix, Alix-NT, Alix-CT, Alix ⌬PGY, or Alix ⌬PP14 were run on SDS-PAGE gel and, after transfer to a nylon membrane, overlaid with recombinant Myc-ALG-2. Bound ALG-2 was revealed with an anti-Myc monoclonal antibody. B, expression levels of each transfected Alix mutant was controlled by using a polyclonal anti-FLAG antibody.
FIG. 7. Activation of cell death by Alix as well as the neuroprotective effect of Alix-CT requires the ALG-2 binding site. After 4 DIV, CGN were transfected with an expression vector coding for GFP, Alix, Alix lacking either endophilin (Alix ⌬PP14), ALG-2 (Alix ⌬PGY) binding domains, Alix-CT, Alix-CT deleted of its ALG-2 binding domain (Alix-CT ⌬PGY), or ALG-2. One day later, cultures were either kept in K25 medium (A) or deprived of potassium (K5 medium; B) and grown for 24 h. Cultures were then fixed, and transfected neurons were revealed by immunofluorescence using the appropriate anti-tag antibody. The transfected neurons showing condensed nuclei (Hoechst staining) were scored as dying, and the percentage of transfected dying neurons was determined. For each experiment, a total of 200 -400 transfected neurons per construct were counted in 10 random fields in two or three different wells; each experiment was performed at least three times. overexpressing neurons. We noticed that CGN containing Caspa-Tag TM inside Alix-CT vacuoles had a condensed nucleus, suggesting that, when detectable in our system, the caspases had already overwhelmed the defense system. These observations are reminiscent of those made in Alzheimer's disease and Down's syndrome brains, where neurons from affected regions show cytoplasmic granulovacuoles believed to be of autophagic origin and which are filled with activated caspase-3 (50)(51)(52).
In agreement with other reports (31-33), we have here observed that the death of CGN induced by potassium deprivation is mediated by both caspase-dependent and -independent cell death, because a pancaspase inhibitor was only able to block nuclear fragmentation without impairing cell death. Therefore, the impairment of caspase activation seen in neurons overexpressing Alix-CT cannot explain all of the beneficial rescuing effect of the truncated protein. Here again, the rescuing effect of Alix-CT was shown to be highly dependent upon the integrity of its ALG-2-binding site. Our demonstration that, besides playing a role in endosome trafficking, Alix also regulates caspase-independent neuronal death in conjunction with the calcium-binding protein ALG-2 may open new avenues toward the understanding of the mechanisms that allow cell destruction in the absence of caspase activity.