Increased internalisation and degradation of GLT-1 glial glutamate transporter in a cell model for familial amyotrophic lateral sclerosis (ALS)

Amyotrophic lateral sclerosis (ALS) is an adult-onset, chronic neuromuscular disorder characterised by the selective degeneration of cortical and spinal/bulbar motor neurons. Approximately 10% of cases are hereditary (familial ALS), 15-20% of which are associated with mutations in the gene Cu²⁺-Zn²⁺ superoxide dismutase (SOD1). Among the various mechanisms proposed as initiators or potentiators of neuronal damage in ALS (reviewed in Julien, 2001; Cleveland and Rothstein, 2001), glutamate excitotoxicity and oxidative damage seem to be correlated and mutually reinforcing.

Glutamate is the primary excitatory neurotransmitter in the mammalian CNS, and so-called glutamate `excitotoxicity' is due to excess extracellular glutamate, which can induce neuronal degeneration in vivo and in vitro (Lipton and Rosenberg, 1994; Rothstein, 1996). Glutamate is not metabolised in the extracellular environment, and the maintenance of normal glutamatergic neurotransmission, or the prevention of glutamate-induced neurodegenerative disorders, depends on the presence of active glutamate transport systems in glial cells and neurons. Among the glutamate (or excitatory amino acid) transporters, EAAC1 (EAAT3), EAAT4 and EAAT5 are neuronal (Kanai and Hediger, 1992; Fairman et al., 1995; Arriza et al., 1997), whereas GLT-1 (EAAT2) and GLAST (EAAT1) are predominantly astroglial (Storck et al., 1992; Pines et al., 1992). The astroglial GLT-1 and GLAST are the most important isoforms involved in keeping extracellular glutamate levels below potentially excitotoxic concentrations (Rothstein, 1996). A number of studies have documented the loss of GLT-1 protein in the motor cortex and spinal cord of patients with sporadic or familial ALS (Rothstein et al., 1995; Bristol and Rothstein, 1996), and abnormal glutamate metabolism (Rothstein et al., 1990).

The hypothesis that oxidative damage plays a role in the pathogenesis of ALS is supported by the discovery that 1-2% of cases are due to mis-sense mutations in the SOD1 gene encoding an enzyme that ordinarily scavenges the free radicals produced in response to oxidative stress (Rosen et al., 1993). More than 100 distinct SOD1 gene mutations have been identified, most of which involve the substitution of a single amino acid and retain full activity. It has been widely shown that the mutant enzyme causes selective neuronal degeneration due to acquired toxicity, rather than the loss of enzymatic function. The toxic functions of SOD1 mutants are probably due to misfolding (Khare et al., 2003; Lindberg et al., 2002), and misfolded mutants can catalyse aberrant reactions, thus leading to increased oxidative conditions in the cell (Wiedau-Pazos et al., 1996; Yim et al., 1996; Bogdanov et al., 1998), and/or function as aggregation centers in which essential molecules are sequestered (Johnston et al., 2000; Okado-Matsumoto and Friedovich, 2002). Both of these mechanisms may account for the increased oxidative stress and damage documented in the tissues of ALS transgenic mice (Ferrante et al., 1997; Hall et al., 1998).

Increasing evidence indicates that ALS is a consequence of reciprocal interactions between multiple cell types rather than a cell autonomous disease, although the hallmark of both sporadic (sALS) and familial ALS (fALS) is the highly selective degeneration of motor neurons (Pramatarova et al., 2001; Lino et al., 2002; Clement et al., 2003). The fact that the exclusively neuronal expression of mutant SOD1 associated with fALS is not sufficient to cause the disease in mice, and also that non-neuronal cells such as astrocytes and microglia are involved in a vicious cycle, suggest that extraneuronal toxic molecules produced by these cells and/or cell damage may critically contribute to the disease (reviewed by Rao and Weiss, 2004). In this context, the specific loss of the astroglial GLT-1, which has been documented in animal models of ALS expressing SOD1 mutants (Bruijn et al., 1997; Bendotti et al., 2001), is highly relevant to the pathogenesis of ALS. One open question is whether the loss of the transporter is a consequence of the neurodegeneration induced by mutant SOD1 or whether the presence of mutant SOD1 protein in astrocytes is sufficient to affect the expression of GLT-1.

The mechanism underlying the downregulation of GLT-1 is unknown. No genomic mutations in GLT-1 have been found, and the overall levels of GLT-1 mRNA are normal in ALS patients and SOD1 mutant mice (Bristol and Rothstein, 1996; Bendotti et al., 2001). It has been shown that the oxidative reactions catalysed by mutant SOD1 impair GLT-1 transport (Trotti et al., 1999), but it is not known whether or how these conditions cause the loss of GLT-1 protein.

In order to test whether the presence of mutant SOD1 in astrocytes is sufficient to affect the expression of GLT-1 and to study the molecular mechanisms of this regulation, we have developed and characterised a cell system consisting of epithelial Madin-Darby Canine Kidney (MDCK) cell lines that stably express high levels of human SOD1 G93A (G cell lines). Glial GLT-1/EAAT2 or GLAST/EAAT1, neuronal EAAC1/EAAT3 and chimeric GLT-1 and EAAC1 glutamate transporters were transiently transfected in these and other MDCK cell lines stably expressing wild-type SOD1 (S cell lines), and their expression and distribution were analysed by biochemical and morphological assays.

The cDNAs of human wild-type or mutant SOD1 (carrying a single base mutation that produces a protein with alanine instead of glycine in position 93) were obtained from total RNA of transgenic mice expressing human wild-type or G93A SOD1 by RT-PCR (Bendotti et al., 2001) using the forward primer: 5′-GCCGGATCCTGCAGTCCTCGGAACCAGGA-3′; and the reverse primer, 5′-AAGGAAAGAAGCGGCCGCAGGATAACAGATGAGTTAAGGG-3′. Full-length cDNAs were cloned into the BamHI and NotI sites of the mammalian expression vector pcDNA3 (Invitrogen, Paisley, UK), and the sequence of the resulting clones was confirmed by DNA sequencing (BMR, CRIBI, University of Padova).

The original cDNAs encoding rat GLT-1 (Pines et al., 1992), rabbit EAAC1 (Kanay and Hediger, 1992) and rat GLAST (Storck et al., 1992), kindly provided by B. Kanner (Hebrew University), M. Hediger (Harvard Institute of Medicine) and W. Stoffel (University of Cologne), were subcloned in the mammalian expression vector pCB6 (Brewer and Roth, 1991). A ClaI-XbaI GLT-1 and a Mlu-BamHI EAAC1 fragment obtained from the original cDNAs were subcloned in similarly digested pCB6 mammalian expression vectors. A modified pCB6 vector, in which the original polylinker from restriction site KpnI to BamHI was replaced by the polylinker of pBluescript II SK^– from KpnI to BamHI in order to obtain unique SmaI and EcoRI sites, was used to subclone the NaeI-EcoRI GLAST fragment obtained from the original cDNA. The GLT-EAAC and EAAC-GLT chimeras, were made by fusing the cytoplasmic tail of EAAC1 (residues 433-524) to amino acid 462 of GLT-1, and the tail of GLT-1 (residues 463-573) to amino acid 432 of EAAC1, respectively, using a unique EcoRV restriction site generated by Quickchange point mutation (Stratagene, La Jolla CA): i.e. C 1484 of GLT-1 cDNA was replaced by T, and G 1473 of EAAC1 cDNA by A. The substitution did not change the GLT-1 amino acid sequence, but caused a conservative substitution of valine433 to isoleucine in the EAAC1 amino sequence. The sequence of all the constructs was confirmed by DNA sequencing, and alterations in the cellular localisation of point-mutated EAAC1 were excluded by immunofluorescence staining.

The MDCK (strain II) cells were cultured as previously described (Pietrini et al., 1992) with 2.5×10⁴ cells/cm² seeded on glass coverslips or petri dishes. Twenty-four hours after plating, the cells were transiently transfected or co-transfected by means of CaPO₄-mediated transfection (Pietrini et al., 1994). Two independent clones for each stable recombinant cell line were selected and fully characterised. All of the experiments shown in the figures were performed using clones 6 (G) and 16 (S), but the graphs and statistical analyses were obtained using clones 6 and 7 (G), and clones 14 and 16 (S).

GLT-1 and GLAST expression was assessed using antibodies raised against isoform-specific peptides (Perego et al., 2000). Rabbit polyclonal anti-EAAC1 antibodies were generated against synthetic peptides consisting of amino acids 1-16 at the N-termini and 500-523 at the C-termini of rabbit EAAC1. No cross-reactivity of the EAAC1 affinity-purified antibodies with the endogenous canine EAAC1 was ever observed in the MDCK cells. LIN-7 was detected using rabbit polyclonal antiserum raised against the histidine-LIN7 fusion protein (Perego et al., 2002), and the surface marker E-cadherin was detected using E-cadherin monoclonal antibodies (Sigma). The expression of wild-type and mutant (G93A) human SOD1 was assessed using commercially available sheep anti-SOD1 antibodies (Calbiochem, Schwalbach, Germany). The antibody did not detect the endogenous SOD1 by immunofluorescence and recognised a faint band of lower electrophoretic mobility than transfected human SOD1 in western blots. Commercial antibodies were used to assess the localisation of β-catenin (Sigma, St Louis, MO), early endosome A1 protein (EEA1) (BD Transduction Laboratories, Lexington, KY) and myc 9E10 (Oncogene, Science, Cambridge, MA) for the myc-tagged epithelial GABA transporter lacking the last five amino acids (Δ5BGT) (Perego et al., 1999).

Cells were fixed in 4% paraformaldehyde and permeabilised with 0.5% Triton X-100 (TX-100). Immunostaining with primary antibodies was followed by incubation with rhodamine-conjugated anti-rabbit, FITC-conjugated anti-mouse or anti-sheep antibodies, and CY5-conjugated anti-mouse antibodies from Jackson Immunoresearch (West Grove, PA).

For WGA labelling, the MDCK cell lines stably expressing the G93A mutant of SOD1 (transiently transfected for GLT-1 two days before the experiment) were incubated on ice with FITC-WGA (Sigma) in PBS with 0.1 mM Ca²⁺, 1 mM Mg²⁺ and 1% bovine serum albumin (BSA) for 30 minutes. The cells were washed to remove unbound labelled lectin, and cultured for 1 hour at 37°C before fixation in paraformaldehyde and GLT-1 staining. For lysoTracker labelling, the cells transfected as described above were cultured in medium containing 0.5 mM FITC-lysoTracker (Molecular Probes, Eugene, OR) for 1 hour before fixation and GLT-1 staining. An MRC-1024 laser-confocal scanning microscope (BioRad, Hercules, CA) was used for morphological analysis.

Forty-eight hours after the transient transfection of GLT-1 in the MDCK cell lines, the cells were biotinylated with NHS-biotin (Pierce Chemical Co., Rockford, IL) and lysed (Sargiacomo et al., 1989). In order to biotinylate lateral surface markers, the cells were pretreated with 5 mM Na-EDTA to open the tight junctions. The biotinylated surface proteins were allowed to bind to streptavidin beads (Sigma) for 2-3 hours and were then separated from the supernatant by centrifugation. The beads contained surface proteins and the supernatant intracellular proteins. After washing, the proteins were released from the beads by incubation with SDS loading buffer, separated by 10% SDS-PAGE and transferred onto nitrocellulose (Shleicher and Shull, Dassel, Germany). The blots were probed for the expression of transporters, E-cadherin, SOD1 and LIN-7 using the primary antibodies described above, with anti IgG or protein A conjugated to peroxidase as the secondary reagent, and visualised by ECL (Perkin-Elmer Life Science, Boston, MA). The signal intensity was densitometrically quantified using NIH Image 1.59 software. Statistical significance was determined using Student's t-test.

To inhibit the internalisation of surface transporters, the cells were cultured for 30 minutes in hypertonic medium consisting of MEM supplemented with 0.45 M sucrose. Protein synthesis was inhibited by culturing the cells in culture medium supplemented with 60 μg/ml cycloheximide (CHX, Sigma) for 6 hours. Lysosomal acidification and lysosomal protease activity were blocked by culturing the cells for 4-12 hours with regular medium supplemented with 100 μM chloroquine (CHQ, Sigma). After treatment, cells were collected and total protein determined in a standard manner before western blot analysis.

It has been shown that expression of the SOD1 mutant selectively impairs the activity of and decreases protein levels in the astroglial GLT-1 isoform of the glutamate transporter in animal models of ALS expressing mutated SOD1 (Bendotti et al., 2001; Howland et al., 2002). One remaining question is whether loss of the transporter is a consequence of the neurodegeneration induced by mutant SOD1 or whether the presence of mutant SOD1 in astrocytes is sufficient to affect the expression of GLT-1. In order to study the cell-autonomous molecular mechanisms underlying SOD1 mutant-mediated GLT-1 regulation, we generated and characterised MDCK cell lines stably expressing wild-type SOD1 (S clones) or mutant G93A (G clones), which is one of the best characterised and widely used human ALS-associated SOD1 mutations. MDCK cells were chosen because they are well characterised, can be used for biochemical and morphological studies and have been used to study the regulation of many transporters. The toxicity of mutant SOD1 is dose-dependent and stable MDCK cell lines continuously express it at high levels (tenfold more than in other cell lines) that are similar to those of the endogenous wild-type in neuronal cells. Moreover, the SOD1 antibody does not immunofluorescently cross-react with the canine isoform, and so immunofluorescence can be used to characterise the clones.

For the purposes of this study, we used clones 6 and 7 expressing the G93A mutant and clones 14 and 16 expressing wild-type SOD1. Nearly all cells in the clones expressed the transfectants but at heterogeneous levels, and the level of expression of the G93A mutant in the G MDCK cell lines was never comparable with wild-type SOD1 in the S cells (Fig. 1,). Various hypotheses may explain the lower level of expression in clones 6 and 7: higher levels of the mutant causing cell death, and thus enabling the selection of these cells, greater susceptibility of the mutant to degradation owing to its incorrect folding (Lindberg et al., 2002), and/or a higher affinity of the antibody for the wild-type compared to the mutant enzyme. The selected G and S MDCK clones or parental MDCK cell lines (M) were transiently transfected with the cDNA encoding the glial GLT-1 or the neuronal EAAC1 isoform of the glutamate transporter gene family, and their total expression was analysed by western blotting. Immunostaining for GLT-1 indicated its markedly reduced expression in G clones, whereas the presence of the SOD1 mutant had no effect on the expression of EAAC1 (Fig. 1A-C). The selectivity of the G93A toxic effect on GLT-1 was further verified by transfecting the GAT-1 neuronal isoform of the GABA transporter family: the SOD1 G93A mutant did not alter the expression of the neuronal GAT-1.

As the carboxy-terminal region of GLT-1 is responsible for inhibiting transport activity in oocytes expressing ALS-linked SOD1 mutations (Trotti et al., 1999), we tested whether it might make the neuronal EAAC1 isoform sensitive to the toxic action of the G93A mutant. Western blotting revealed reduced levels of the EAAC1 with the cytosolic tail of GLT-1 (EAAC-GLT) but not of a GLT-1 having the tail of EAAC1 (GLT-EAAC) (Fig. 1A-C), which suggests that mutant SOD1 may reduce the expression of GLT-1 by acting on a specific tail sequence.

To answer the question whether mutant SOD1 has to be stably expressed to affect GLT-1 expression, GLT-1 was transiently co-transfected with wild-type or mutant SOD1 with no difference in GLT-1 expression being observed (Fig. 1D), thus indicating that, in order to exert its toxic effect on GLT-1 expression, the SOD1 G93A mutant must be continuously expressed in MDCK cells.

Taken together, the results shown (Fig. 1A,D) suggest that toxic molecules accumulated in cells stably expressing G93A may act on a GLT-1 tail sequence. As the amino acids sequences that are possible targets of the oxidative reactions catalysed by the G93A mutant are exclusively contained in the carboxy terminal of the GLT-1 isoform, we exposed G cells to antioxidant and oxidant molecules in order to prevent or increase downregulation of GLT-1 in G cells. Under our experimental conditions, the total expression of GLT-1 was not affected by these treatments (Fig. 1E).

MDCK cells are the most thoroughly characterised of the polarised cell lines. They have two morphologically and functionally distinct plasma membrane domains: the apical free surface characterised by microvilli, and the lateral junctional surface connecting adjacent cells. As the apical and basolateral plasma membrane domains have distinct transporter isoform compositions (Muth and Caplan, 2003), we investigated whether the downregulation of GLT-1 was due to domain mis-targeting.

The S and G MDCK clones were transiently transfected with cDNAs encoding glutamate transporters, and 48 hours later, fixed and processed by means of triple immunofluorescence staining. Confocal analysis revealed a strikingly different distribution of GLT-1 between the cells expressing wild-type SOD1 and those expressing the G93A mutant. Like the fully polarised MDCK cell lines stably expressing GLT-1 (Cheng et al., 2002) (G.P., unpublished observations), the S clones showed apical and lateral GLT-1 distribution: its apical distribution was revealed by the red staining on the apical free surface (arrow), and its lateral distribution by colocalisation with β-catenin (Fig. 2A,B). In the G93A-expressing cells, GLT-1 (apical and lateral) surface localisation was less evident, whereas there was clear intracellular perinuclear staining (Fig. 2A,B). The G93A-dependent intracellular redistribution was specific for GLT-1, as the apical localisation of EAAC1 and the basolateral localisation of GLAST were not affected by the expression of wild-type or mutant SOD1 (Fig. 2B). In the cell lines expressing wild-type SOD1 (S clones), the apical and lateral distribution of the chimeric EAAC-GLT transporter was similar to that of GLT-1, thus confirming a reported observation (Cheng et al., 2002) of a targeting signal for the apical surface in the cytosolic tail of EAAC1. The chimeric transporter was predominantly localised on the cell surface, but intracellular perinuclear staining was also observed in G cells (Fig. 2A,B), which may account for the western blot data indicating that the presence of specific sequence in the tail of GLT-1 is responsible for the decreased GLT-1 levels in these cells.

These findings also suggest that G93A selectively affects the surface expression of GLT-1, and that the effect does not depend on the cell-surface domain in which the transporters are located.

We next quantified the GLT-1 surface/intracellular ratio in surface biotinylation experiments, and found that approximately 75-85% of the total transporter was biotinylated in the parental and SOD1-expressing MDCK cell lines, whereas the percentage of biotinylated GLT-1 was less than 50% in the G93A-expressing cell lines (Fig. 3A,B). Similar levels of E-cadherin were biotinylated, thus indicating lateral surface availability to the biotin reagent and the generally unchanged surface composition of the cell lines. As expected for a protein distributed at the cell surface and intracellularly, the ratio of biotinylated GLT-1 in G cells compared to that in M or S cells was much lower than that between total transporter expressions (Fig. 3C). These data support the intracellular relocalisation of GLT-1 and a decreased surface expression of GLT-1 in G cell lines.

An intracellular localisation can be a consequence of GLT-1 accumulation within the exocytic or endocytic pathways of G93A-expressing cell lines. Double immunofluorescence staining with endoplasmic reticulum and Golgi apparatus markers excluded the accumulation of GLT-1 in these organelles (data not shown), whereas histochemical assays indicated that GLT-1 was localised within the endocytic compartments in the G93A clones. In a morphological endocytosis assay, surface glycoproteins were labelled with FITC-wheat germ agglutinin (WGA) at 0°C, and chased at 37°C for 60-90 minutes before fixation and staining for GLT-1. It was found that internalised WGA-labelled glycoproteins accumulated in the intracellular structures containing GLT-1 (Fig. 4), thus indicating that these are in endocytic compartments.

In order to investigate the endocytic localisation of GLT-1 in the G clones, we analysed the effect of endocytosis inhibition on GLT-1 distribution. It has been widely reported that hypertonic sucrose (0.45 M) treatment specifically inhibits endocytosis through clathrin-coated pits (Hansen et al., 1993; Heuser and Anderson, 1989), and we found that a 30-minute treatment caused the disappearance of intracellular staining, with GLT-1 fully colocalising with β-catenin, a marker of the lateral junctional surface (Fig. 4B). These results indicate that GLT-1 reaches the cell surface where it is then quickly removed and that the internalised transporter accumulates within the endocytic pathway in G93A-expressing cell lines.

Having demonstrated that the ablation of endocytosis prevents the removal of the transporter from the plasma membrane and its accumulation along endocytic pathways, we investigated whether internalised GLT-1 recycles to the cell surface or is routed to the degradative pathway. GLT-1 recycling to the plasma membrane can occur rapidly at the level of the early endosomes located in the cell periphery (sorting endosomes) or slowly from the recycling endosomal compartment located in the cell perinuclear region and containing both apical and laterally internalised proteins (Gruenberg and Maxfield, 1995; Mellman, 1996). To distinguish these possibilities, we used double immunofluorescence staining with antibodies against early-endosome antigen 1 (EEA1), and GLT-1 co-transfected with Δ5BGT, a truncated version of the BGT1 epithelial GABA transporter that accumulates in the slow recycling compartment of MDCK cells (Perego et al., 1999).

Confocal analysis revealed no colocalisation of the intracellular transporter with EEA1 or Δ5BGT (Fig. 5), but in vivo labelling with FITC-lysotracker (Molecular Probes) revealed the colocalisation of GLT-1 with the degradative acidic compartments (Fig. 5). These results indicate that the intracellular staining of GLT-1 is not due to its accumulation in recycling compartments; the entry/exit ratio is therefore close to 1 in early endosomes and more than 1 in the lysotracker-labelled compartment. Lysotracker accumulates along the acidic degradative pathway, which is formed by heterogeneous structures, and imperfect colocalisation of GLT-1 with this marker suggests the localisation of the two proteins in sub-compartments of this pathway.

To investigate directly whether G93A induces changes in the degradation rate of the transporter, protein synthesis was inhibited and the total expression of the transporter analysed. The degradation of GLT-1 was clearly increased, as demonstrated by the 50% reduction in GLT-1 levels after cycloheximide treatment in these, but not in M or S cells. To demonstrate that GLT-1 protein is downregulated because of the increased degradation of surface GLT-1, we determined the levels of GLT-1 after inhibiting the activity of the degradative acidic compartments with chloroquine (CHQ). In line with the degradation of the transporter after its removal from the plasma membrane, overnight incubation with chloroquine had a larger effect on GLT-1 level in G cells. A twofold increase of GLT-1 expression in M cells after the overnight inhibition of lysosomal degradation is consistent with the half-life of the transporter (Fig. 6A,B).

The loss of the astroglial GLT-1 glutamate transporter has been widely documented in ALS patients and animal models expressing human SOD1 mutations. As soluble neuronal factors dependent on neuronal differentiation and activity are required to sustain the level of GLT-1 expression in glial cells (Perego et al., 2000), the loss of GLT-1 could merely be a consequence of motor neuron degeneration; however, the fact that decreased GLT-1 levels have been documented before neuronal damage in transgenic rats expressing high levels of mutated SOD1 (Howland et al., 2002) suggests that GLT-1 downregulation may be controlled by both neuronal and autonomous cell factors driven by SOD1 mutations. A cell autonomous regulation of GLT-1 mediated by mutant SOD1 is further suggested by data showing that the function of GLT-1 is impaired by mutant SOD1 in the presence of free radicals (Trotti et al., 1999).

In this study, we documented an autonomous cell mechanism mediated by G93A affecting GLT-1 expression. The continuous expression of the G93A mutant of SOD1 in MDCK cell lines causes loss of the transporter by increasing its internalisation and degradation. In addition, we found that transporter targeting to degradation requires the cytosolic tail of GLT-1.

The cell system consisting of MDCK cells stably expressing SOD1 G93A reproduces the specific downregulation of the astroglial GLT-1 glutamate transporter associated with ALS-linked SOD1 mutations. Western blot and immunofluorescence data indicate the decreased expression and intracellular relocalisation of GLT-1, but not decreased expression of the neuronal glutamate EAAC1 transporter and no intracellular relocalisation of the EAAC1 and GLAST isoforms. Further confirmation of the specificity of the effect was obtained by expressing a membrane carrier belonging to the GABA family (the neuronal GAT-1), whose expression was never significantly reduced in the cells expressing the SOD1 G93A mutant.

The altered GLT-1 expression and distribution in G93A-expressing cells are closely related events, because intracellular GLT-1 accumulation is a consequence of the increased internalisation of the surface transporter ultimately targeted to degradation in lysosomal compartments. The first indication that GLT-1 accumulates in the endocytic degradative compartments of MDCK-G93A cells came from the histochemical endocytosis assay. Intracellular GLT-1 colocalised with internalised WGA-labelled glycoproteins when the cells were chased in regular medium for more than 1 hour after WGA labelling. Colocalisation of the lysosomal marker cathepsin D and WGA-labelled glycoproteins indicated that, after this period of chase, WGA-labelled glycoproteins localise in lysosomal compartments (data not shown). Double staining of the cells to verify the lysosomal localisation of GLT-1 was impaired because the available cathepsin D for staining in MDCK cells and GLT-1 antibodies are both polyclonal rabbit antibodies.

In line with GLT-1 colocalisation in endocytic degradative compartments and the increased rate of degradation, cycloheximide treatment for 6 hours greatly reduced the expression of GLT-1 in G, but not in M and S cells. The GLT-1 degradation during this period was therefore below the level of detection in these clones, and mutant SOD1 affects the expression of the transporter by acting downstream its synthesis.

Confirmation of the endocytic nature of these structures and the lysosomal degradation of internalised GLT-1 came from experiments in which a hypertonic medium (sucrose 0.45 M) was used to inhibit endocytosis (Hansen et al., 1993; Heuser and Anderson, 1989). As expected in the case of any internalised protein targeted to degradation, inhibition of endocytosis led to the disappearance of intracellular GLT-1 staining (Fig. 4), thus suggesting that, after internalisation, the transporter is quickly degraded in G93A cells. When the cells were treated with chloroquine overnight to inhibit the activity of the acidic degradative compartments, GLT-1 levels increased in G cells (∼threefold) and in MDCK cells (∼twofold). A twofold increase in MDCK cells is consistent with the transporter half-life, but the threefold increase further supports greater GLT-1 degradation in G cells. In contrast, the small increase in GLT-1 in S cells after chloroquine treatment, together with the generally increased expression of transporters in these cells (Fig. 1) may be explained by a protective effect of the wild-type enzyme on the degradation of all of the analysed transporters, including GLT-1.

To demonstrate that the internalised transporter is routed to lysosomal degradation, we excluded its intracellular accumulation in recycling compartments by means of double-staining experiments with markers of early endosomes (EEA1) and the slow recycling compartments (Δ5BGT). The colocalisation of intracellular GLT-1 with lysotracker (Fig. 5) indicates that GLT-1 was targeted to degradation in acidic pre-/lysosomal compartments.

These data are in line with the hypothesis that post-translational modifications in GLT-1, directly or indirectly induced by SOD1 mutants may cause transporter internalisation and degradation, but the nature of G93A-induced modifications remains unclear.

The fusion of the C-terminal cytosolic tail of GLT-1 to the neuronal glutamate transporter EAAC1 (chimera EAAC-GLT) was sufficient to downregulate the expression of the otherwise unaffected neuronal isoform. However, the accumulation of EAAC-GLT in the endocytic pathway is much less, and we were unable to modulate the expression of the chimeric transporter with cycloheximide and chloroquine treatments (S.M. and G.P., preliminary results), thus suggesting that other GLT-1 domains may participate in its G93A-mediated regulation. Previous data have demonstrated that the cytosolic tail of GLT-1 is responsible for inhibiting transport activity in oocytes expressing ALS-linked SOD1 mutations (Trotti et al., 1999), and our data suggest that this domain is also involved in the degradation of the transporter.

The two main hypotheses so far proposed in order to explain mutant SOD1-mediated toxicity in ALS are aberrant oxidative damage and protein aggregation (Wiedau-Pazos et al., 1996; Cleveland and Liu, 2000; Cleveland and Rothstein, 2001; Okado-Matsumoto and Fridovich, 2002). These may not be mutually exclusive because an enhanced production of oxygen radicals could also modulate and facilitate the formation of mutant protein aggregation and cause cell toxicity. GLT-1 internalisation and degradation may therefore be triggered by the oxidation of residues located in the GLT-1 tail and/or the SOD1 aggregate trapping of a specific protein whose binding to the GLT-1 tail is required to keep the transporter on the cell surface, as has been shown in the case of the BGT1 epithelial GABA transporter (Perego et al., 1999).

It has been shown that the reactive oxygen species produced in motor neurons induce oxidation and disrupt glutamate uptake (Rao et al., 2003), and that increased oxidation in oocytes expressing SOD1 mutations can selectively affect the transport activity of GLT-1, and that this inhibitory effect requires the cytosolic C-terminal domain of GLT-1 (Trotti et al., 1999). However, it is not clear whether oxidative reactions catalysed by mutant SOD1 also affect the level of GLT-1. Unlike the oocyte findings, the toxic effect of the SOD1 mutant on GLT-1 did not require the addition of oxidant compounds such as H₂O₂ or MQ in the G clones. One possible explanation for this discrepancy is that the continuous expression of the G93A mutant makes an increase in cellular oxidative conditions unnecessary. However, there was no effect on the expression of GLT-1 in G cells exposed to antioxidant reagents (2-ME, NAC) and no signs of oxidative stress or functional and morphological mitochondrial alterations (G.P., unpublished data) in these cells. The lack of involvement of oxidative stress in the downregulation of GLT-1 mediated by the G93A mutant has also been shown recently in primary cultures of cortical astrocytes (Tortarolo et al., 2004). Expression of wild-type and mutant SOD1 for 4 days was sufficient to affect the expression of endogenous GLT-1 in the astrocyte cell system whereas the stable expression of G93A was required to downregulate GLT-1 in MDCK cells. Neither the stable expression of wild-type SOD1 nor transient co-expression of GLT-1 and SOD1 G93A for 2 days had any detectable effect on GLT-1 levels. It therefore seems that the MDCK cell system requires continuous expression of high levels of the SOD1 mutant to recapitulate the cell modifications leading to transporter loss in glial cells.

Cytosolic C-terminal domains are potential targets of various post-translational modifications that may cause the internalisation of plasma membrane proteins, including phosphorylation. It has been shown that increased endocytosis of GLT-1 and other transporters is due to kinase-dependent regulation (Daniels and Amara, 1999; Kalandadze and Robinson, 2002), and that the activities and protein levels of various kinases are altered in the brain and spinal cord of transgenic mice overexpressing mutant SOD1 genes, as well as in ALS patients (reviewed by Hu and Krieger, 2002). Identifying the amino acid residues involved in the transporter loss will help to clarify the molecular mechanism underlying the toxic functions of mutant SOD1 in ALS.

In conclusion, our data indicate a cell-autonomous toxic effect of SOD1 G93A that causes a specific post-translational downregulation of GLT-1 by increasing the removal of the transporter from the cell surface and targeting the internalised transporter for degradation.

We thank Drs B. Kanner, J. Handler and W. Stoeffel for providing reagents. We also thank Dr N. Borgese and C. Bendotti for their comments on the manuscript and K. Smart for his help in preparing the text. Special thanks to N. Borgese for her support and encouragement. This research was supported in part by Ministry of Health grant IRCCS-2002, Fondazione Mondino, to N. Borgese and G.P., and by ALS Association, grant 2001 to G.P.