Alix [ALG-2 (apoptosis-linked gene 2)-interacting protein X] is a ubiquitinous adaptor protein first described for its capacity to bind to the calcium-binding protein, ALG-2. Alix regulates neuronal death in ways involving interactions with ALG-2 and with proteins of the ESCRT (endosomal sorting complex required for transport). Even though all Alix interactors characterized to date are involved in endosomal trafficking, the genuine function of the protein in this process remains unclear. We have demonstrated recently that Alix and ALG-2 form in the presence of calcium, a complex with apical caspases and with the endocytosed death receptor TNFR1 (tumour necrosis factor α receptor 1), thus suggesting a molecular coupling between endosomes and the cell death machinery.
Numerous observations suggest that, in several neurodegenerative diseases, including ALS (amyotrophic lateral sclerosis), AD (Alzheimer's disease) and Niemann–Pick disease, affected neurons display an early impairment in the endosomal system. For example, in AD brains, endosome abnormalities appear in neurons long before amyloid plaque and neurofibrillary tangle formation . The endosomal system is composed of a series of intracellular compartments within which endocytosed molecules traffic. Most of the endocytosed proteins return to the membrane, while some meant for degradation are selectively entrapped in vesicles budding from the membrane into the lumen of endosomes. This process leads to the formation of endosome intermediates called MVBs (multivesicular bodies), filled with intraluminal vesicles; these and their cargoes will be hydrolysed after fusion of the MVBs with lysosomes . One sorting signal used for the trafficking through MVBs is the mono-ubiquitination of cytosolic parts of transmembrane receptors, which occurs after their activation at the cell surface. Once inside endosomes, ubiquitinated cytoplasmic domains pointing out into the cytosol trigger the sequential building of the ESCRTs (endosomal sorting complexes required for transport) on to the endosomal membrane. CHMPs (charged multivesicular body proteins) of ESCRT-III associate to form a lattice entrapping transmembrane proteins and which is necessary for endosomal membrane vesiculation . Another actor of the vesicle budding inside endosomes might be the protein Alix [ALG-2 (apoptosis-linked gene 2)-interacting protein X]/AIP1 (actin-interacting protein 1), which binds to both ESCRT-I and ESCRT-III, as well as to LBPA (lysobisphosphatidic acid), a lipid facilitating the budding of vesicles inside MVBs .
An important hint that MVB abnormalities lead to neurodegenerative diseases came from the discovery that Niemann–Pick type C disease is linked to mutations inducing abnormal accumulation of unesterified cholesterol in MVBs and the concomitant death of neurons . More recently, Reid et al.  found that the spastin protein encoded by SPAST, mutations of which were known to cause degeneration of cortical motoneurons in hereditary spastic paraplegia, interacts with CHMP1B of ESCRT-III. Mutations of the ESCRT-III subunit CHMP2B have also been reported to be linked with cases of FTD (frontotemporal dementia), the second most common cause of presenile dementia, characterized by a severe cortical atrophy, and with non-SOD1 (superoxide dismutase 1) ALS, associated with degeneration of motor neurons [7,8]. These genetic data fit with the morphological data to suggest that endosomal function is central to neurodegenerative pathologies. The connection between endosomal dysfunction and neurodegeneration means that certain trafficking and/or processing events taking place in endosomes are critically important for neuronal homoeostasis. One possibility is that molecular interactions within endosomes control the initiation of a neuronal death programme. The adaptor protein Alix seems to mediate interactions of this type.
Alix ESCRTs endosomes
We first characterized Alix while searching for proteins capable of binding the calcium-binding protein ALG-2 . At that time, the aim of our research was the definition of molecules involved in death induced by intracellular calcium, and our interest for ALG-2 began as it was reported to be required for T-cell apoptosis . Alix is a 90 kDa cytosolic protein with no enzymatic signature, but with a long proline-rich C-terminal region, which binds ALG-2 only when the latter is complexed to calcium. Other Alix interactors have a demonstrated role in endocytosis: CIN85 (Cbl-interacting protein of 85 kDa)/SETA [SH3 (Src homology 3) domain-containing gene expressed in tumorigenic astrocytes]/Ruk (regulator of ubiquitous kinase), first involved in endocytosis of ubiquitinated tyrosine kinase receptors ; endophilins A, regulating clathrin-dependent endocytosis of surface receptors ; Tsg101 (tumour susceptibility gene 101) and SNF7/CHMP4B, members of ESCRT-I and -III respectively [13,14]; and LBPA, a phospholipid which is highly concentrated inside MVBs and may promote intraluminal vesiculation of endosomes . Alix might orchestrate the deformation and fission of lipid bilayers since it was described as being involved in the budding of vesicles inside MVBs [15,16] and the abscission reactions that complete mammalian cell division . The role played by the protein in deforming membranes was recognized as central by enveloped viruses, which recruit it in order to bud off membranes . Surprisingly, however, neither we nor other laboratory groups have found any significant effect of Alix on endocytosis and degradation of EGF (epidermal growth factor) or transferrin receptors in mammalian cells ; thus the final consequence of Alix activity in MVBs remains largely unclear.
Alix links endosomes to neuronal death
One hint that Alix is involved in neuronal death comes from our observations of its up-regulation in rat-degenerating neurons of the striatum in a model of Huntington's disease  or of the hippocampus during epileptic seizures. In this latter case, neuronal death is known to occur because of calcium entering through glutamate receptors and is accompanied by a massive up-regulation of endocytosis and of autophagy . Alix up-regulation could be a cause of neurodegeneration, since its overexpression by transfection was sufficient to activate caspases and induce death of post-mitotic cerebellar granule neurons. On the other hand, expression in these cultured neurons of the Alix C-terminal half (Alix-CT) of the protein, which acts as a dominant-negative mutant, blocked death induced by potassium deprivation. Both the pro- and anti-apoptotic effects of Alix and Alix-CT respectively required the integrity of the ALG-2-binding site, suggesting that the calcium-dependent interaction of ALG-2 with Alix is instrumental in controlling neuronal death . We need here to emphasize that the cerebellar neurons used in these experiments were post-mitotic and therefore that the effects seen on apoptosis cannot be due to the role of Alix in cytokinesis. Furthermore, the neurons were cultured in the absence of serum and of exogeneous factors and only survive because of chronic depolarization obtained by high extracellular potassium concentrations. This rules out that, in this paradigm, Alix mutants act on cell death by modifying endosomal trafficking of trophic receptors. We went on to test whether Alix/ALG-2 might play a role in cell death in a more physiological setting, which is the programmed cell death occurring naturally during development of the chick embryo. Using electroporation, we found that expression of some mutants of Alix blocked death of motor neurons during normal development. This blocking effect was tightly dependent on binding to both ALG-2 and ESCRT-I and -III. Our interpretation of these results is that some truncated forms of Alix behave as dominant-negative mutants inhibiting the formation of an ALG-2–Alix–ESCRT complex necessary for cell death . Therefore the Alix–ALG-2 complex could make a link between endosomes and a signalling, or an execution step of neuronal death .
Alix–ALG-2: a new caspase-activating platform bound to ESCRT?
Using immunoprecipitations coupled to MS, we found that, in dying cerebellar neurons, Alix associates in a complex containing the apical caspase 8. We have reproduced the formation of this complex in BHK (baby-hamster kidney) cells and shown that it is tightly dependent on calcium and on the capacity of Alix to interact with ALG-2 . Another apical caspase, caspase 9, also co-immunoprecipitated with Alix and ALG-2, whereas execution caspases 3 and 7 did not. Caspases are synthesized as zymogens, and activation occurs through caspase proteolytic cleavage. Intrinsic and extrinsic pathways of apoptosis led to the recruitment of apical caspases on to multimeric complexes, inducing proximity and thereby autocatalytic activation of the zymogens. Activated apical caspases in turn cleave and thereby activate execution caspases, which dismantle the cell undergoing apoptosis. We therefore hypothesize that Alix functions as a scaffold allowing autocatalytic activation of caspase 8 and 9 and that the build-up of the complex is controlled by calcium binding to ALG-2. In agreement with this, expression of Alix-CT or knockdown of Alix protected BHK cells from death induced by thapsigargin, which elevates cytosolic calcium by blocking the sarcoplasmic/endoplasmic reticulum Ca2+-ATPase (F. Strappazzon, S. Torch, C. Chatellard-Causse, A. Petiot, B. Blot, J.-M. Verna and R. Sadoul, unpublished work).
Our results, together with the tight relationship of Alix with endosomes, suggest a coupling of apical caspases to these latter organelles. A possible understanding of this unexpected link came from the work of Schneider-Brachert et al.  who reported that, in lymphocytes, activation of caspase 8 by TNFR1 (tumour necrosis factor α receptor 1) occurs only after the receptor has been endocytosed and passed on to MVBs. This finding has introduced the notion of death-inducing signalling endosomes . We recently demonstrated that Alix binds to endocytosed TNFR1, whereas ALG-2 binds to pro-caspase 8 . In the presence of calcium, the complex made by Alix and ALG-2 contains both pro-caspase 8 and TNFR1 and might thus be instrumental in adapting the caspase zymogen on to TNFR1-containing receptors. The physiological relevance of these interactions was given by the observation that expression of AlixΔALG-2, which interacts with TNFR1, but not with caspase 8, blocks TNFR1-induced cell death without interfering with endocytosis of the receptor. Furthermore, we observed that the same mutant blocked programmed cell death of developing motor neurons in the chick embryo, which is controlled by TNFR1 .
Alix controls caspases…and what else?
Even though the idea of an Alix–ALG-2 complex allowing activation of apical caspases on the surface of endosomes is appealing, it cannot entirely explain the mechanisms by which Alix mutants block neuronal death. Indeed, we found that expression of P35, a baculovirus pan-caspase inhibitor, delays, but does not block, death of motor neurons in vivo, in contrast with Alix mutants which allow long-term survival of the same neurons . This is in good agreement with numerous publications that have reported that, in most cases, inhibiting or knocking out caspases only blocks nuclear destruction, but does not allow neuronal survival. Thus neuronal death is the sum of caspase-dependent mechanisms involved in the dismantling of the nucleus and of caspase-independent mechanisms allowing destruction of the cytoplasm. The capacity of Alix mutants to block cell death both in vivo and in vitro suggests that the protein drives both caspase-dependent and -independent processes.
Caspase-independent mechanisms are ill-defined, but the presence of autophagic vacuoles in dying cells in which caspases are inhibited, or in certain neurons dying during development and in adult central nervous system, has led several authors to postulate that macro-autophagy is a caspase-independent mechanism contributing to cell destruction [26,27]. Macro-autophagy, hereafter referred to as autophagy, is the major pathway for degradation of long-lived proteins and the only known pathway for elimination of organelles. This process begins with the formation of double-membrane vacuoles, referred to as autophagosomes, which engulf cytoplasmic material. Autophagosomes fuse with endosomes, thereby giving rise to amphisomes, which in turn fuse with lysosomes . Several proteins involved in the biogenesis of MVBs, among which the ATPase SKD1 (suppressor of K+ transport defect 1), as well as the three ESCRTs, modulate autophagy [29,30]. With this in mind, we have challenged the hypothesis that Alix might drive caspase-independent cell death by controlling autophagy. Using cerebellar granule neurons, we found that Alix-induced cell death is accompanied by a drastic increase in the number of autophagosomes, whereas expression of Alix-CT reduced the formation of autophagosomes in dying neurons. However, in cell lines whose viability is not influenced by Alix, overexpression or down-regulation of the protein had no effect on autophagosomes or autophagic protein degradation induced by amino acid depletion . These results show that, in non-dying cells, Alix does not participate directly in autophagy, and reinforce the conclusion of other studies demonstrating that its role differs from that of classical ESCRT proteins. Thus, even though Alix's role in apoptosis seems to be due to a control of caspase activation by both calcium and endosomes, its role in controlling caspase-independent processes remains obscure.
ESCRTs: from Cell Biology to Pathogenesis: Biochemical Society Focused Meeting held at Robinson College, Cambridge, U.K., 26–28 August 2008. Organized and Edited by Katherine Bowers (University College London, U.K.), Juan Martin-Serrano (King's College London, U.K.) and Paul Whitley (Bath, U.K.).
ALG-2-interacting protein X
C-terminal half of Alix
amyotrophic lateral sclerosis
charged multivesicular body protein
endosomal sorting complex required for transport
TNFα receptor 1
This work was supported in part by Inserm, the University Joseph Fourier, grants from the Association Française contre les Myopathies [grant number MNM 2007], the Association pour la Recherche sur la Sclérose Latérale Amyotrophique [grant number 060711] and the Association pour la Recherche contre le Cancer [grant number 4948]. A.-L.M.-M. was supported by a fellowship from the Association Française contre les Myopathies.
Present address: Experimental Medicine and Toxicology, Imperial College London, London W12 0NN, U.K.
Present address: European Center for Brain Research, Santa Lucia Foundation, Molecular Neuroembryology Unit, 00143 Rome, Italy
Present address: MRC Functional Genetic Unit, Department of Physiology, Anatomy and Genetics, University of Oxford, Oxford OX1 3QX, U.K.