|Home | About | Journals | Submit | Contact Us | Français|
Author contributions: A.S., J.S.M., and D.G.C. designed research; A.S. and J.S.M. performed research; R.L.W., and J.M.S. contributed unpublished reagents/analytic tools; A.S., J.S.M., and D.G.C. analyzed data; A.S., J.S.M., J.S.D., and D.G.C. wrote the paper.
GLT-1, the major glutamate transporter in the adult brain, is abundantly expressed in astrocytic processes enveloping synapses. By limiting glutamate escape into the surrounding neuropil, GLT-1 preserves the spatial specificity of synaptic signaling. Here we show that the amyloid-β peptide Aβ1–42 markedly prolongs the extracellular lifetime of synaptically released glutamate by reducing GLT-1 surface expression in mouse astrocytes and that this effect is prevented by the vitamin E derivative Trolox. These findings indicate that astrocytic glutamate transporter dysfunction may play an important role in the pathogenesis of Alzheimer's disease and suggest possible mechanisms by which several current treatment strategies could protect against the disease.
Glutamate transporters bind and remove glutamate from the extracellular space. In adult hippocampus this task is performed primarily by two astrocytic transporter subtypes: GLT-1 and GLAST (homologous to human excitatory amino acid transporters EAAT1 and EAAT2, respectively) (Kanai and Hediger, 1992). The activity of these transporters is crucial to ensure pathway specificity of synaptic transmission and plasticity (Diamond and Jahr, 1997; Tsvetkov et al., 2004), prevent glutamate-induced excitotoxicity (Tanaka et al., 1997), regulate ammonia detoxification (Felipo and Butterworth, 2002), and modulate activity-dependent glucose utilization (Voutsinos-Porche et al., 2003). Several studies indicate that glutamate transporters are impaired in Alzheimer's disease (AD) (Masliah et al., 1996; Lauderback et al., 2001; Jacob et al., 2007). GLT-1 loss and reduced detergent solubility of the remaining GLT-1 occur in early prodromal AD and become more prominent with increasing cognitive impairment (Woltjer et al., 2010). Partial GLT-1 loss in an animal model of AD also provokes early-occurring cognitive deficits (Mookherjee et al., 2011), suggesting that GLT-1 loss is capable of driving cognitive impairment in the context of Aβ-related neuropathology. Whether Aβ-related pathogenic processes have a functional impact on the rate at which astrocytes remove endogenous, synaptically released glutamate is unknown. Here we show that Aβ1–42 induces rapid GLT-1 mislocalization and internalization in astrocytes, which leads to a marked reduction in the rate at which astrocytes clear up synaptically released glutamate from the extracellular space. Importantly, we show that pretreatment with the vitamin E derivative, Trolox, prevents the effects of Aβ1–42 on GLT-1 and astrocytic glutamate clearance.
P14–P21 C57BL/6 mice of either sex were deeply anesthetized with halothane and decapitated, in accordance with the NINDS and Veterans Affairs Animal Use Committees guidelines. Transverse hippocampal slices (250 μm thick) were obtained with a vibrating blade microtome (VT1000S, Leica Microsystems). The slicing solution, kept at 4°C and continuously bubbled with a mixture of 95% O2-5% CO2, contained the following (in mm): 119 NaCl, 2.5 KCl, 0.5 CaCl2, 1.3 MgSO4 · 7H2O, 4 MgCl2, 26.2 NaHCO3, 1 NaH2PO4, 22 glucose, 320 mOsm, pH 7.4. After cutting, slices were kept in this solution in a submersion chamber at 34°C for 30 min, and at room temperature thereafter for up to 5 h. Recording solution contained the following (in mm): 119 NaCl, 2.5 KCl, 2.5 CaCl2, 1.3 MgSO4 · 7H2O, 26.2 NaHCO3, 1 NaHPO4, 22 glucose, 300 mOsm, pH 7.4, saturated with 95% O2-5% CO2. All recordings were performed at 34–36°C. The following drugs were routinely added to the artificial CSF to block AMPA, NMDA, and metabotropic glutamate receptors, GABAA, GABAB, and adenosine receptors: 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzoquinoxaline-7-sulfonamide disodium salt (NBQX; 10 μm), (R,S)-3-(2-carboxypiperazin-4-yl)-propyl-1-phosphonic acid (CPP; 10 μm), 2-[(1S,2S)-2-carboxycyclopropyl]-3-(9H-xanthen-9-yl)-d-alanine (LY341495; 1 μm), (R,S)-α-methylserine-O-phosphate (MSOP; 100 μm), picrotoxin (100 μm), 3-[[(3,4-dichlorophenyl)methyl]amino]propyl]diethoxymethyl)phosphinic acid (CGP52432; 5 μm), 8-cyclopentyl-1,3-dipropylxanthine (DPCPX; 1 μm). Astrocytes in CA1 stratum radiatum were identified under IR-DIC using an upright microscope (Axioskop 2FS, Zeiss). Electrophysiological recordings were made with 2.5–3.0 MΩ patch pipettes filled with (in mm): 120 CsCH3SO3, 10 EGTA, 20 HEPES, 2 MgATP, 0.2 NaGTP, 5 QX-314Br, biocytin (0.4% w/v) 290 mOsm, pH 7.2. After each recording, the slices were fixed overnight in 4% paraformaldehyde and stored in PBS for histochemical processing. Transporter currents were evoked by applying 50 μs, constant voltage, electrical pulses to stratum radiatum through bipolar stainless steel electrodes (Frederick Haer Company) located ~100 μm away from their cell body. Single and paired pulses, 100 ms apart, were delivered every 10 s. Reagents were from Sigma and Tocris Bioscience. Currents were filtered at 2 kHz, digitized at 5 kHz, and acquired with custom-made software (A.S., J.S.D) written in IgorPro (Wavemetrics). Data analysis was performed within the IgorPro environment using custom-made software (A.S.). The analysis was performed on transporter currents isolated by subtracting the average response to single pulse stimulations from that evoked by paired stimuli (Diamond, 2005; Scimemi et al., 2009). The time course of the glutamate clearance waveform (F(t)) was measured as the centroid <t, where:
Changes in t reflect changes in the rising and decaying phase of glutamate clearance.
C57BL/6;129Sv mice of either sex were used in accord with protocols approved by the University of Washington Animal Care and Use Committee. Neocortex from neonatal mice was digested with papain and grown in MEM supplemented with 5% fetal calf serum, GlutaMAX, 20 mm glucose, and penicillin/streptomycin (Life Technologies). Confluent astrocyte cultures were purified by shaking, replated, grown 7–10 d, and treated with 0.15 mm dibutyryl-cAMP (Sigma), 10 μm Ara-C, 3% FBS, GlutaMAX in MEM for 96 h to increase GLT-1 expression (Swanson et al., 1997). Medium was replaced with normal medium for 6 h before treatment with Aβ1–42 (30 min, 37°C). While chilled at 4°C, astrocytes were treated with Pierce EZ-Link Sulfo-NHS-SS-Biotin (Thermo Scientific) according to the manufacturer's protocol.
Synthetic human Aβ1–42 peptide (CPC Scientific) oligomerization was performed as described previously (Stine et al., 2003) and confirmed using Western blots (Fig. 1D) from 16.5% tricine-gel (Bio-Rad) SDS polyacrylamide gels immunoblotted with 6E10 antibodies (Covance). Control condition did not contain Aβ1–42 peptide but was otherwise identically prepared. As a final step, 0.1% (w/v) BSA (used as a carrier protein) was added to each aliquot and mixed before storage. Equal volumes of Control and Aβ1–42 aliquots were diluted into either electrophysiology solutions or astrocyte medium (depending on the experiment) to a final concentration of 500 nm Aβ1–42 and 0.0005% BSA protein carrier.
Slices containing biocytin-filled astrocytes were fixed in 4% paraformaldehyde and permeabilized with 0.25% Triton X-100/PBS. Biocytin and GLT-1 were visualized with streptavidin Alexa Fluor 488 and anti-rabbit/Alexa Fluor 568 (Life Technologies), respectively. Slices were imaged on a Nikon A1R confocal microscope. Each stack contained 15–30 images acquired at 0.25 μm intervals. Final figures were prepared using Adobe Photoshop, where only linear brightness and contrast adjustments were performed identically on Aβ1–42-treated and control slice images.
Protein lysates from slices and cultured astrocytes were prepared using 1% Triton X-100 buffer [in mm: 20 Tris, 1 EDTA, 0.5 EGTA, 250 sucrose, Protease Inhibitor Cocktail (EMD Millipore)]. Lysate protein concentrations were determined by bicinchoninic acid assay (Thermo Scientific). Lysates were electrophoresed on 4–20% SDS polyacrylamide gels (Bio-Rad) and transferred onto Immobilon-P membranes (Millipore). Antibodies recognizing GLT-1 (Millipore) and AB12 (gift from Dr. David Pow, RMIT University of Melbourne, Australia) (Mookherjee et al., 2011), GLAST (Cell Signaling Technology), and EAAC1 (Alpha Diagnostics) were used. Pyruvate kinase (Rockland) and Sypro Ruby protein blot stain (Life Technologies; conducted before blocking as per manufacturer's protocol) were used as controls for soluble and insoluble protein fractions, respectively. Bands were visualized with ECL-Plus (GE Healthcare). Optical densitometry measurements were performed with ImageJ (NIH), and protein bands were normalized with respect to their respective protein load signals (pyruvate kinase for Triton X-100-soluble protein fractions and Sypro Ruby for detergent-insoluble protein fractions) and normalized against identically determined average control values. Biotinylated cell surface proteins were pulled down from 10 μg of whole-cell Triton X-100 lysates using Neutravidin beads (Thermo Scientific). Because intracellular GLT-1 expression predominates in purified cultured astrocytes, pull-downs were run alongside 1 μg of their corresponding lysates to avoid saturating the linear range of the imaging device (GE Healthcare ImageQuant LAS/4000). The GLT-1 detergent insolubility measures were performed on acute brain slices prepared and stored as described for the electrophysiological experiments. Slices were treated for 30 min with vehicle control or 500 nm Aβ1–42, with or without a 1 h pretreatment with 100 μm Trolox (Sigma). Slices were extracted three times (1 h each) in 1%Triton X-100 lysis buffer and clarified by high-speed centrifugation (100,000 × g) to separate the Triton X-100-soluble supernatant and Triton X-100-insoluble pellet. For ubiquitin/GLT-1 immunoprecipitation/Western blots, 100 μg samples (obtained from the first of three serial Triton X-100 protein extractions prepared during the GLT-1 detergent insolubility experiments; see Fig. 3) were immunoprecipitated with anti-Ubiquitin P4D1 monoclonal antibody (Santa Cruz Biotechnology) and Protein G Agarose beads (Life Technologies).
Student's t tests and ANOVA were performed as appropriate (Winer et al., 1991) using SPSS software (IBM).
We recorded synaptically activated transporter currents (STCs) from astrocytes in acute mouse hippocampal slices, before and after applying a subsaturating concentration of the broad-spectrum glutamate transporter antagonist TBOA (dl-threo-β-benzyloxyaspartic acid; 10 μm). By monitoring the effect of TBOA on the STCs, we estimated the time course of synaptically released glutamate clearance from astrocytes as previously described (Diamond, 2005; Scimemi et al., 2009) (Fig. 1A; Materials and Methods). To test whether Aβ1–42 alters glutamate clearance in astrocytes, we applied this analytical approach to STCs recorded from astrocytes in control slices (i.e., slices perfused with 0.0005% bovine serum albumin) and in slices treated with monomeric or oligomeric Aβ1–42 (500 nm) and 0.0005% bovine serum albumin protein carrier. TBOA reduced the STC amplitude to the same extent in all conditions (normalized STC amplitude: Control 0.25 ± 0.08, n = 5; Aβ1–42 oligo 0.27 ± 0.04, n = 9; Aβmono 0.32 ± 0.08, n = 8; pControl-Aβ1–42 oligo = 0.80, pControl-Aβ1–42 mono = 0.56). However, as shown in Figure 1A–C, the time course of glutamate clearance (expressed as the centroid, t) was significantly slower in slices treated with oligomerized or monomeric Aβ1–42 (tclearance Control = 4.1 ± 0.5 ms, n = 5; tclearance Aβ1–42 oligo = 8.3 ± 0.5 ms, n = 9, *pControl-Aβ1–42 oligo = 0.044; tclearance Aβ1–42 mono = 7.1 ± 0.9 ms, n = 8, *pControl-Aβ1–42 mono = 0.036). These results suggest that Aβ1–42 rapidly impairs astrocytic glutamate transport. Because there was no difference between the effects of monomeric or oligomeric Aβ1–42 preparations on glutamate clearance (p = 0.47), in all remaining experiments we used oligomerized Aβ1–42 preparations (hereafter referred to as Aβ1–42), as they contained a mixture of monomeric and higher-order Aβ species potentially emulating a spectrum of synthetic Aβ1–42 conformers similar to those found in AD patients (Fig. 1D; Materials and Methods).
The slower time course of glutamate clearance from astrocytes treated with Aβ1–42 could be mediated by one or more astrocytic glutamate transporter subtypes. We asked whether GLT-1 and/or GLAST mediated the effects of Aβ1–42. Under our experimental conditions, GLAST contributes ~1/3 of the total astrocytic glutamate transport, with the rest contributed by GLT-1 (Scimemi et al., 2009). We found that the specific GLT-1 inhibitor DHK (dihydrokainate; 300 μm) prolonged the time course of glutamate clearance (tclearance DHK = 10.0 ± 2.6 ms, n = 5; Fig. 1E,F), in agreement with previous reports (Diamond, 2005; Scimemi et al., 2009). The time course of glutamate clearance in the presence of DHK was comparable to that observed in slices treated with Aβ1–42 (p = 0.53) and, in the presence of DHK, Aβ1–42 failed to prolong glutamate clearance (tclearance DHK+Aβ1–42 = 9.3 ± 2.0 ms, n = 6; p = 0.85). These results indicate that the Aβ1–42-induced slowing of glutamate clearance could be accounted for entirely by effects on glutamate uptake by GLT-1.
One possible mechanism underlying the effects of Aβ1–42 on astrocyte-mediated glutamate clearance is reduced membrane expression of GLT-1. To test this hypothesis, we used confocal microscopy to image biocytin-filled astrocytes and GLT-1. Normally GLT-1 is expressed only at the plasma membrane of astrocytes (Chaudhry et al., 1995). In keeping with this, GLT-1 immunoreactivity (red) appeared to localize primarily at or near the cell surface of control astrocytes filled with biocytin (green) (Fig. 2A, top row; single z-plane image showing one of 19 biocytin-labeled astrocytes examined). In contrast, GLT-1 immunoreactivity appeared to colocalize extensively with the biocytin signal in the cytoplasm of Aβ1–42-treated astrocytes (Fig. 2A, bottom row; single z-plane image showing one of 9 biocytin-labeled astrocytes examined). This finding suggests that Aβ1–42 alters GLT-1 expression at or near the astrocytic plasma membrane.
To test quantitatively whether Aβ1–42 induced GLT-1 internalization, we turned from acute hippocampal slices to cultured astrocytes, which are better suited for measuring cell-surface protein expression via cell-surface biotinylation. Repeating the Aβ1–42 treatment protocol on cultured astrocytes (i.e., 30 min application of Aβ1–42, 500 nm), we found that GLT-1 cell surface expression was significantly reduced by Aβ1–42 treatment (mean optical density: Aβ1–42 = 0.67 ± 0.07, n = 8; Control = 1.0 ± 0.09, n = 8, p = 0.01; Fig. 2B,E), while GLAST cell surface expression was not altered (Aβ1–42 = 1.04 ± 0.19, n = 8; Control = 1.0 ± 0.19, n = 8, p = 0.86, Fig. 2D,E). Total protein levels of GLT-1 and GLAST in cultured astrocytes were not significantly altered by Aβ1–42 exposure (total mean GLT-1: Aβ1–42 = 0.99 ± 0.11, n = 8; Control = 1.0 ± 0.04, n = 8, p = 0.933; total mean GLAST: 0.90 ± 0.11, n = 8, Control = 1.0 ± 0.08, n = 8, p = 0.50; Fig. 2C,E). Similarly, in hippocampal slices, the total levels of GLT-1, GLAST, and the neuronal glutamate transporter EAAC1 were unaffected by Aβ1–42 treatment (GLT-1: Control = 1.0 ± 0.07, n = 6, Aβ1–42 = 0.86 ± 0.08, n = 6; p = 0.20; GLAST: Control = 1.0 ± 0.11, n = 6, Aβ1–42 = 0.77 ± 0.08, n = 6; p = 0.12; EAAC1: Control = 1.0 ± 0.11, n = 6, Aβ1–42 = 0.86 ± 0.10, n = 6; p = 0.39; Western blots not shown). Together, these data indicate that Aβ1–42 slowed clearance of synaptically released glutamate by disrupting GLT-1 localization at/near the astrocytic plasma membrane.
Previous work showed that Aβ1–42 oxidatively damages GLT-1 (Guo and Mattson, 2000; Lauderback et al., 2001) and that oxidative stress promotes formation of aberrant high molecular weight GLT-1 complexes that are insoluble in detergents such as Triton X-100 (Haugeto et al., 1996; Trotti et al., 1997, 1998). These findings raise the possibility that Aβ1–42 alters GLT-1 surface expression and astrocytic glutamate clearance by activating oxidative stress responses. To test this hypothesis we assessed the ability of Aβ1–42 to induce GLT-1 detergent insolubility in hippocampal slices and tested whether Trolox, a highly water-soluble derivative of the anti-oxidant, vitamin E (Quintanilla et al., 2005), was capable of blocking the effects of Aβ1–42 on GLT-1.
We found that GLT-1 levels in the Triton X-100-insoluble protein fraction (i.e., proteins remaining as a pellet after three serial Triton X-100 extractions) from Aβ1–42-treated slices were significantly elevated compared with controls (Fig. 3A,B, left). By contrast, Aβ1–42 treatment had no effect on GLT-1 levels in the detergent-soluble protein fraction (Fig. 3B, right). Normal levels of detergent-insoluble GLT-1 were restored by pretreating the slices with Trolox (100 μm; 1 h preincubation plus the 30 min Aβ1–42 exposure period) (F(2,74) = 13.605, p < 0.001; Fig. 3B, left).
Aberrant post-translational modifications are closely associated with detergent-insoluble protein aggregation in AD (Woltjer et al., 2005). In keeping with this, Aβ1–42 promotes pathologic oxidative stress-induced covalent modifications of numerous proteins and lipid moieties (Sultana et al., 2006). Intriguingly, recent studies found that ubiquitination of GLT-1 causes it to internalize from the cell surface of astrocytes (Martínez-Villarreal et al., 2012). These findings raised the question whether oxidative stress elicited by Aβ1–42 exposure induces GLT-1 ubiquitination. To address this possibility, hippocampal slices were treated as described above. Ubiquitinated proteins were immunoprecipitated with anti-Ubiquitin antibodies and Western blotted for GLT-1. The results in Figure 3, C and D, show that Aβ1–42 treatment significantly increased GLT-1 ubiquitination and that this effect was blocked by Trolox pretreatment (F(2,17) = 5.354, p < 0.016).
To address the functional significance of Trolox pretreatment on GLT-1 function in astrocytes, we again recorded STCs from astrocytes in hippocampal slices (Fig. 4). We first confirmed that the Trolox pretreatment on its own did not change the time course of glutamate clearance compared with control conditions (tTrolox = 5.8 ± 0.5 ms, n = 10 p = 0.09). Notably, the time course of glutamate clearance was not prolonged by Aβ1–42 in slices pretreated with Trolox (tAβ1–42+Trolox pretreat = 4.7 ± 0.7 ms, n = 12,). By contrast, the time course of glutamate clearance was still prolonged by Aβ1–42 and Trolox in slices not pretreated with Trolox (tAβ1–42+Trolox no pretreat = 7.6 ± 0.9 ms, n = 11) (F(2,30) = 4.086, p < 0.027). These results suggest that the ability of Trolox to prevent the effects of Aβ1–42 is not due to direct molecular interactions between Trolox and Aβ1–42.
The ability of Trolox to prevent the Aβ1–42-induced changes in GLT-1 ubiquitination, membrane expression, detergent insolubility, and transport function suggests that oxidative stress is a relevant injury mechanism impairing the ability of hippocampal astrocytes to carry out their critical neuroprotective function.
Our results show that Aβ1–42 prolongs the lifetime of synaptically released glutamate at astrocytic membranes via oxidative stress, GLT-1 ubiquitination, and mislocalization at the cell membrane. These data support the notion that GLT-1 plays a key role in the pathogenesis of AD (Masliah et al., 1996; Lauderback et al., 2001; Jacob et al., 2007; Woltjer et al., 2010; Mookherjee et al., 2011) and suggest potential opportunities for therapeutic intervention. Notably, our results show that anti-oxidant pretreatment—not acute exposure—can prevent entirely the effects of Aβ1–42 on GLT-1. Importantly, Aβ1–42 treatment in acute slices recapitulates the aberrant GLT-1 detergent insolubility found in hippocampi of mildly cognitively impaired and late-stage AD patients (Woltjer et al., 2010).
The ability of Aβ1–42 to double the amount of time it takes to clear synaptically released glutamate suggests that GLT-1 dysfunction may facilitate Aβ-related neurotoxicity. It also suggests that even under conditions where extracellular glutamate levels are restored to their normal baseline levels, Aβ1–42 may still promote the spread of glutamate from one synaptic domain to another—an effect that could impair the normal activity of entire neuronal networks.
The effects of Aβ1–42 on GLT-1 mislocalization, protein insolubility, and ubiquitination argue for the functional significance of rapidly occurring post-translational modifications of GLT-1 in AD-related pathogenesis. There is also evidence that longer-term exposure to Aβ1–42 can cause GLT-1 loss in astrocytes that is mediated by numerous inflammation-related transcriptional processes (Abdul et al., 2009). GLAST levels are also reduced by inflammatory transcriptional processes (Sama et al., 2008). This may be related to why total glutamate transporter levels in our experiments trended lower, albeit nonsignificantly, due to short-term Aβ1–42 exposure. Taken collectively, such findings and the data herein suggest that AD-related pathologic processes can disturb astrocytic glutamate transporters via multiple interrelated molecular pathways that ultimately lead to glutamate transporter loss in AD (Masliah et al., 1996; Jacob et al., 2007; Abdul et al., 2009; Woltjer et al., 2010).
The molecular mechanisms underlying the effect of Aβ1–42 on GLT-1 therefore have important mechanistic implications for: (1) development of new treatments to combat oxidative stress in AD, like those pursued in ongoing clinical studies testing the efficacy of combining vitamin E and memantine, a drug thought to suppress excessive glutamate stimulation of the brain (ClinicalTrials.gov # NCT00235716); and (2) development of therapeutic strategies to limit the neurotoxicity of prolonged glutamate exposure/spread in AD (Lipton, 2005).
This work was supported by the Veteran's Affairs Office of Research and Development Medical Research Service (D.G.C.), NIH (Grants T32 AG000258, J.S.M.; and RO1 NS055804, J.M.S.), and NINDS Intramural Research Program (Grant NS002986, J.S.D.). We thank Mike Ahlquist, Ping Zhu, and Ning Li for excellent technical assistance.
The authors declare no competing financial interests.