Displacing hexokinase from mitochondrial voltage-dependent anion channel (VDAC) impairs GLT-1-mediated glutamate uptake but does not disrupt interactions between GLT-1 and mitochondrial proteins

Abstract

The glutamate transporter GLT-1 is the major route for the clearance of extracellular glutamate in the forebrain, and most GLT-1 protein is found in astrocytes. This protein is coupled to the Na⁺-electrochemical gradient, supporting the active intracellular accumulation of glutamate. We recently used a proteomic approach to identify proteins that may interact with GLT-1 in rat cortex, including the Na⁺/K⁺-ATPase, most glycolytic enzymes, and several mitochondrial proteins. We also showed that most GLT-1 puncta (~70%) are overlapped by mitochondria in astroglial processes in organotypic slices. Based on this analysis, we proposed that the glycolytic enzyme hexokinase 1 (HK1) might physically form a scaffold to link GLT-1 and mitochondria because HK1 is known to interact with the outer mitochondrial membrane protein, voltage-dependent anion channel (VDAC). In the present study, we first validated the interactions between HK-1, VDAC and GLT-1 using forward and reverse immunoprecipitations. We also provided evidence that a subfraction of HK1 co-localizes with GLT-1 in vivo. We found that a peptide, known to disrupt the interaction between HK and VDAC, did not disrupt interactions between GLT-1 and several mitochondrial proteins. In parallel experiments, we found that displacement of HK from VDAC reduced GLT-1-mediated glutamate uptake. These results suggest that although HK1 forms co-immunoprecipitatable complexes with both VDAC and GLT-1, it does not physically link GLT-1 to mitochondrial proteins. However, the interaction of HK1 with VDAC supports GLT-1-mediated transport activity.

Introduction

Glutamate is the principal excitatory neurotransmitter in the CNS. Synaptic concentrations of glutamate are maintained at low levels (approximately 25 nM) against brain concentrations that approach 10 mmol/Kg in order to ensure proper signaling and prevent excitotoxicity (Herman and Jahr 2007; Schousboe 1981). Glutamate is cleared from the synapse by a family of Na⁺-dependent glutamate transporters of which GLT-1 and GLAST, the astroglial glutamate transporters, predominate (for reviews, see Danbolt 2001; Robinson 1999). These transporters couple the uptake of one molecule of glutamate with the co-transport of 3 Na⁺ ions and a H⁺, followed by the counter-transport of a K⁺ ion (Zerangue and Kavanaugh 1996).

Using mass spectrometry, we previously identified subunits of the Na⁺/K⁺-ATPase, most of the glycolytic enzymes, and several mitochondrial proteins in GLT-1 immunoprecipitates from rat cortical homogenates (Genda et al. 2011). In this same study, we also showed that GLT-1 and mitochondria overlap more than 70% of the time in astroglial processes in cultured organotypic hippocampal slices and that this overlap was greater than would be observed by chance, consistent with a physical interaction between GLT-1 and mitochondria. We proposed that this complex might provide metabolic support to facilitate glutamate uptake. We also proposed that the glycolytic enzyme HK1 might serve as a scaffolding protein linking plasma membrane GLT-1 with mitochondrial proteins. A known interaction between HK and the plasma membrane glucose transporter GLUT4 gave precedence to the possibility of direct HK interactions with GLT-1 (Zaid et al. 2009).

Hexokinase catalyzes the first and one of the rate limiting steps of glycolysis, the phosphorylation of glucose to form glucose-6-phosphate. Glucose phosphorylation also represents the first step in the synthesis of glycogen and the pentose phosphate shunt. In brain, most HK1 is bound to mitochondria by a hydrophobic N-terminal domain that interacts with the voltage dependent anion channel (VDAC) found in the outer mitochondrial membrane (Linden et al. 1982; Nakashima et al. 1986 Sui and Wilson 1997). VDAC interacts with the adenine nucleotide translocase (ANT) in the inner mitochondrial membrane (Beutner et al. 1998 Crompton et al. 1998) and many of the mitochondrial proteins form supercomplexes (Acin-Perez et al. 2008). Hexokinase/VDAC interactions link glycolysis and oxidative phosphorylation. HK1 binding of VDAC is thought to increase the catalytic efficiency of both processes by facilitating mitochondrial ATP release from VDAC for glucose phosphorylation and by channeling ADP into mitochondria for oxidative phosphorylation (BeltrandelRio and Wilson 1991; Rodrigues-Ferreira et al. 2012).

Given the importance of glutamate signaling and the metabolic costs associated with maintaining glutamate uptake, it is not surprising that glutamate uptake is intimately associated with metabolism. Glutamate uptake and Na⁺/K⁺-ATPase activity are fueled by glycolysis, glycogenolysis, and oxidative phosphorylation (Fernandez-Moncada and Barros 2014; Genda et al. 2011; Sickmann et al. 2009). Glutamate uptake stimulates glycolysis (Pellerin and Magistretti 1994), glucose uptake (Hamai et al. 1999; Loaiza et al. 2003), and glycogen formation (Hamai et al. 1999; Swanson et al. 1990). In addition, glutamate can itself be oxidized within the mitochondria, potentially fueling its own uptake (Sonnewald and McKenna 2002). In fact, inhibitors of glutamate dehydrogenase (GDH) block glutamate uptake (Whitelaw and Robinson 2013).

In the present study, we validated GLT-1 interactions with VDAC and HK1 using both forward and reverse immunoprecipitations. We then tested whether HK1 might provide a mechanism to link the plasma membrane protein, GLT-1, with mitochondrial proteins (e.g. VDAC, UQCRC2, ANT). In addition, we determined if the interaction between HK1 and VDAC contributes to supporting glutamate uptake.

Methods

All procedures involving the use of animals were reviewed and approved by the Institutional Animal Care and Use Committee of the Children’s Hospital of Philadelphia (Philadelphia, PA USA).

Preparation of tissue/immunoprecipitation

Cortical tissue was harvested from adult male Sprague Dawley rats after euthanasia by decapitation to avoid potential effects of anesthetic agents (Huang and Zuo 2005). Total tissue lysates were prepared as previously described (Genda et al. 2011). Tissue was homogenized in ice cold immunoprecipitation buffer (18.4 ml per gram wet weight), containing 150 mM NaCl, 1 mM ethylene diamine tetraacetic acid (EDTA), 100 mM Tris HCl, pH 7.4, 1% Triton X-100, and 1% sodium deoxycholate plus protease and phosphatase inhibitors (1µg/ml leupeptin, 250 µM phenylmethylsulfonyl fluoride (PMSF), 1 µg/ml aprotinin, 1 mM iodoacetamide, 10 mM NaF, 30 mM sodium pyrophosphate and 1 mM sodium orthovanadate) using a Dounce Teflon/glass homogenizer (7 strokes at 400 RPM). All subsequent steps were performed keeping the tissue at 4° C. Homogenates were rotated on a shaker for 1 hr and then cleared of cellular debris by centrifugation at 13,000 g for 30 min. One ml of lysate was pre-cleared with 80 µl protein-A agarose beads (Invitrogen, Carlsbad, CA) for 1 hr followed by centrifugation (16,500 g for 15 min). After analyses of protein (bicinchoninic acid protein assay kit; Pierce, Rockford, IL), an aliquot of the supernatant containing 500 µg of protein was mixed overnight with either 15µg antibody or a species-matched IgG (mouse IgG was from Zymed Laboratories/Life Technologies, Grand Island, NY, rabbit IgG was from Invitrogen, Carlsbad, CA). Protein complexes were batch extracted using protein-A agarose beads (30 µl) with gentle mixing for 2 hrs. Agarose beads were washed four times in immunoprecipitation buffer before elution of bound proteins by the addition of 25 µl SDS-PAGE loading buffer followed by incubation at 25°C for 45 min or 95°C for 5 min. Incubation at 25°C for 45 min reduces the formation of GLT-1 aggregates and was used for Figure 1B and C (upper panel) and Figure 4B.

Figure 1.

GLT-1, HK1, and VDAC form immunoprecipitatable interactions within the cortex. Anti-GLT-1 (A), anti-HK1 (B), anti-VDAC (C) antibodies or IgG were used for immunoprecipitations from rat cortical lysates (500 µg protein). Immunoprecipitation of the target protein was confirmed in every immunoprecipitation. Data are representative of at least 3 independent experiments.

Figure 4.

HK1-VDAC disruption does not affect the interaction between GLT-1 and mitochondrial proteins, but does reduce sodium-dependent glutamate uptake in crude synaptosomes. Representative western blots of GLT-1 (A) and ANT (B) IPs under control and HK-peptide conditions labeled for GLT-1 and UQCRC2, or ANT and GLT-1, respectively. Uptake of radiolabeled glutamate following treatment with a control or HK-peptide is expressed as percent of control (C). Values are displayed as mean ± SEM; n=3; *p<0.05.

Crude synaptosomal membranes (P2) were prepared as previously described (Robinson et al. 1991) except that the final P2 preparation was resuspended in 4.5 vols. of sucrose (0.32 M). An aliquot (450 µl) of this P2 suspension (containing ~ 1.8 mg of protein) was combined with 50 µl Hexokinase II VDAC binding domain peptide (fused to Antennapedia homeo-domain) (Calbiochem #376816) or control Antennapedia homeo-domain peptide (Calbiochem #287895) that had been resuspended in 0.32 M sucrose (100 µM final concentration) for 30 minutes at 37°C. An aliquot (65 µl) of this material was placed on ice for measurement of uptake (see below). The remainder of the material was centrifuged at 20,000 × g for 20 min. Except for the analyses of HK1 and VDAC interactions, the pellet was resuspended in 1 ml of immunoprecipitation buffer and processed as described above with 200 µl of pre-cleared lysate containing ~240 µg of protein. To test for HK1-VDAC interactions, pre-cleared lysates were incubated overnight using a Pierce crosslink immunoprecipitation columns (Thermo Scientific, Waltham, MA) prepared using 15 µg HK1 antibody or mouse IgG.

Western Blot Analysis

Proteins were separated using 10% SDS-polyacrylamide gels and transferred to polyvinylidene fluoride membranes (PVDF-FL; Millipore, Billerica, MA), and blocked for 1 hr at 25°C in TBS-T (50 mM Tris, 150 mM NaCl, pH 8.0, 0.1% Tween) containing 5% nonfat dry milk. Membranes were then probed with the appropriate primary antibody: rabbit or mouse anti-GLT-1 (1:5,000; courtesy of Dr. J. Rothstein, Johns Hopkins University, Baltimore, MD USA) (Rothstein et al. 1994); mouse anti-HK1 (1:500; Sigma, St. Louis, MO, #WH0003098M1); mouse anti-ubiquinol-cytochrome-c reductase complex core protein 2 (UQCRC2) (1:5000; Abcam, Cambridge, UK, #AB14745), rabbit anti-VDAC (1:1000; Abcam #AB15895); goat anti-ANT (1:50; Santa Cruz, Santa Cruz, CA, #SC9299). Membranes were incubated with fluorescent dye-conjugated anti-rabbit, anti-mouse, or anti-goat secondary antibodies (1:10,000; LiCor Biosciences, Lincoln, NE). Protein bands were visualized using an Odyssey Infrared Imager (LiCor Biosciences).

Glutamate uptake in P2 membrane fraction

Sodium-dependent transport of L-[³H]-glutamate was measured as previously described (Robinson et al. 1991; Whitelaw and Robinson 2013). HKII or control treated synaptosomes (10 µL, containing ~40 µg of protein) were incubated with 0.5 µM L-[³H]-glutamate (PerkinElmer, Waltham, MA USA) in uptake buffer, containing 5 mM Tris base, 10 mM HEPES, 140 mM NaCl, 2.5 mM KCl, 1.2 mM CaCl₂, 1.2 mM MgCl₂, 1.2 mM K₂HPO₄, 10 mM glucose (pH=7.2) at 37°C for 3 min. We have previously determined that ³H-glutamate uptake is linear until at least 5 min in this system (Robinson 1991). Uptake was measured in the absence and presence of sodium by substituting equimolar amounts of choline chloride for NaCl. The assay was terminated by the addition of 2 ml of ice-cold choline-containing buffer. Following termination, the suspensions were filtered onto glass filter paper (FP-100; Brandel, Gaithersburg, MD USA) using a Brandel cell harvester and rinsed three times with 2 ml of cold choline-containing buffer. Radioactivity was solubilized with 5 ml of Cytoscint ES (MP Biochemicals, Solon, OH USA) and measured using scintillation spectrometry (Beckman-Coulter Instruments LS 6500). Sodium dependent uptake was calculated as the difference between the signal in the sodium-containing buffer from the signal in the choline-containing buffer.

Immunofluorescence

Adult rats (10–12 weeks) were anesthetized with isoflurane and transcardially perfused with ice-cold phosphate buffered saline (PBS) followed by 4% paraformaldehyde in PBS, pH 7.4. Brains were post-fixed overnight (4°C), equilibrated in 30% sucrose, flash frozen in isopentane (−50°C), and stored at −80°C. Sections were cut on a freezing-sliding microtome and processed as free-floating sections. Sections were blocked in PBS containing 5% normal serum and 0.4% Triton X-100 for 1 hr and then incubated with rabbit anti-GLT-1 (Rothstein, 1:100) and/or mouse anti-HK1 (Sigma, St. Louis, MO; 1:100) overnight at 4°C. After 3 rinses, they were incubated in secondary antibody (either Alexa Fluor 488 or 633; 1:400, Invitrogen) overnight at 4°C. Sections were mounted on pre-coated slides, cover-slipped with mounting media (VectaShield), and stored at 4°C until analysis. Controls for each experiment included incubations to confirm the species specificity of secondary antibodies and to confirm that signal was dependent on the presence of primary antibodies. Sections were visualized on a Fluoview 1000 confocal microscope (Olympus, Center Valley, PA) equipped with a PlanApo 60× objective (numerical aperture = 1.4). All images were collected in sequential scan mode to minimize cross-contamination of fluorophores.

Intensity correlation analysis / intensity co-localization quotient

Single optical sections were background corrected and filtered using a Gaussian filter (r = 1). All image analysis was conducted using NIH ImageJ software (http://rsb.info.nih.gov/ij/). Images were automatically thresholded and the Intensity Co-localization Quotient (ICQ) values calculated using the intensity correlation analysis (ICA) plug-in to the McMaster Biophotonics Facility ImageJ collection. The ICQs were calculated as described by Li et al (2004); results were compared to zero using a one-sample t test. Images representing the product of the differences from the mean (PDM) intensity values were generated using this plugin.

Results

Interaction between GLT-1, HK1, and VDAC in the cortex

In our earlier study, we identified a number of glycolytic enzymes and mitochondrial proteins in GLT-1 immunoprecipitates from rat brain cortical lysates by mass spectrometry (Genda et al. 2011). Many of these interactions were further validated using conventional immunoprecipitation/western blotting and reverse immunoprecipitation using an antibody against the putative interacting protein. At the time, we were unable to validate the interactions with either HK1 or with VDAC because we could not identify suitable antibodies. For the present study, we found suitable antibodies for this purpose. Using these antibodies, we performed a series of immunoprecipitations followed by western blot analysis to test whether the glial glutamate transporter GLT-1, the glycolytic enzyme HK1, and the mitochondrial protein VDAC interact in rat cortical tissue. Using an anti-GLT-1 antibody for immunoprecipitation, we observe HK1 (Fig 1). In addition, we consistently found ubiquinol-cytochrome-c reductase complex core protein 2 (UQCRC2), as was observed in our earlier study (Genda et al. 2011). In mammals, there are three isoforms of VDAC representing distinct gene products with high homology but different molecular weights due to post-translational modifications (Yamamoto et al. 2006). In the earlier mass spectroscopy analysis, we identified VDAC2 and VDAC3 in GLT-1 immunoprecipitates (Genda et al. 2011). Using an anti-VDAC antibody that recognizes all three isoforms, we observed bands at approximately 29 and 31 kDa in GLT-1 immunoprecipitates. The higher molecular weight band represents the VDAC1 and 2 isoforms, the lower molecular weight band represents VDAC3 (Kerner et al. 2012; Shoshan-Barmatz and Golan 2012; Yamamoto et al. 2006). In these same immunoprecipitates, we also verified the presence of GLT-1 (Fig. 1A, lower panel) that migrates as a monomer with a molecular weight of approximately 66 kDa and a trimer of approximately 200 kDa (Haugeto et al. 1996). These analyses complement our earlier mass spectroscopic analysis providing evidence that the proteins were appropriately identified, but they do not rule out the possibility that the anti-GLT-1 antibodies directly interact with these proteins.

To address this possibility, we performed a series of ‘reverse’ immunoprecipitations to determine if GLT-1 co-immunoprecipitates with these targets using antibodies against these putative interacting proteins. Using an anti-HK1 antibody, we find that UQCRC2, VDAC, and GLT-1 all co-immunoprecipitate with HK1 (Figure 1B, upper panels). Similarly, using an anti-VDAC antibody, we find that GLT-1, HK1, and UQCRC2 all co-immunoprecipitate with VDAC (Figure 1C, upper panels). In both of these sets of immunoprecipitations, we confirmed that the intended target was observed in the immunoprecipitates (Figure 1B &C, lower panels). Although these results do not formally rule-out the possibility that subsets of these interacting proteins interact separately, the simplest explanation is that these proteins participate in a multi-protein complex that includes GLT-1, HK1, UQCRC2, and VDAC.

Previously, we examined the extent of co-localization of mRFP-GLT-1 with HK1-EGFP within the processes of transfected astrocytes in hippocampal slice cultures (Genda et al. 2011). We found that virtually all exogenous HK1 localized to mitochondria. Here, we extended this analysis to look at the extent of co-localization of endogenous HK1 and GLT-1 within rat cortical tissue. We conducted immunofluorescence imaging of HK1 and GLT-1 in sections derived from adult rat cortex (Fig. 2) followed by intensity co-localization analysis. This analysis is based on the premise that if HK1 and GLT-1 interact, then their staining intensities should co-vary. Conversely if the proteins are part of different complexes, then their staining patterns should be segregated. We calculated Intensity Co-localization Quotients (ICQs) for HK1 and GLT-1 in the cortex of rat (Genda et al. 2011; Li et al. 2004). This quotient describes the extent of correlation of the staining intensities for two proteins. If two images have staining patterns that are dependent, then their staining intensities will co-vary around their respective means and the product of their differences from the mean (PDM) values will be positive. The ICQ values are equal to the ratio of number of positive PDM values to the total number of pixels and are distributed between −0.5 and +0.5. Positive ICQ values indicate the relative intensities are correlated and consistent with co-localization. An ICQ value of 0 would indicate random staining, while negative values indicate segregated staining. We see a small, but significant co-variance of signal for GLT-1 and HK1 (ICQ=0.06 ± 0.1, p=0.0006, n=8) within rat cortex. This covariance exists despite the fact that the majority of GLT-1 is found in astrocytes in vivo (Rothstein et al. 1994), while HK1 is present in high amounts in neurons as well (Cahoy et al. 2008). While the majority of GLT-1 expression in cortex is astrocytic, there is also a small pool of GLT-1 in neurons; therefore we cannot rule out the possibility that the observed co-localization is occurring within the neurons (Chen et al. 2004; Furness et al. 2008).

Figure 2.

GLT-1 and HK1 co-localize in rat brain cortex. (A-C) Representative images from adult rat cortex immunostained with antibodies against HK1 (A), GLT-1 (B), and merged (C). (Scale bar=10 µm). (D) Pseudocolor representation of the product displacement of the mean (PDM) values for the image pair (A, B), where pixel intensity is equal to the PDM at that location. Images with positive PDM (dependent) values are displayed in yellow, while those with negative values (segregated) are displayed in violet.

N-terminal HK peptide disrupts HK1-VDAC binding in synaptosomes

For the next series of experiments, we used crude synaptosomal membranes (P2) from cortex so that we could measure Na⁺-dependent L-[³H]-glutamate uptake in parallel. Several lines of evidence indicate that essentially all of the uptake in this preparation is mediated by GLT-1, including the demonstration that the pharmacology of uptake uniquely reflects GLT-1 and genetic deletion of GLT-1 reduces uptake to 5% of control (Arriza et al. 1994; Tanaka et al. 1997, for review see Robinson 1999; Robinson et al. 1991).

We hypothesized that HK1 may act as a cytosolic bridge linking GLT-1 and mitochondria. To test this hypothesis, we used a cell-permeable peptide based on the N-terminal, VDAC-binding domain of HK2 (Pastorino et al. 2002; Sui and Wilson 1997) to disrupt the interaction between HK1 and VDAC (mitochondria), and measured the effect on interactions between GLT-1 and mitochondrial proteins. Several groups have used this peptide to disrupt the interaction between endogenous HK (1 and 2) and VDAC in various cell culture models (Majewski et al. 2004; Pastorino et al. 2002; Sukumaran et al. 2010). The N-terminus of HK2 shares sequence homology (11 of 15 amino acids) and its VDAC binding site with HK1, making the peptide useful for displacing both of these isoforms from VDAC. To confirm that the peptide disrupts the interaction between HK1 and VDAC in the crude synaptosomes, we used an anti-HK1 antibody to immunoprecipitate HK1 and quantified the amount of VDAC immunoreactivity. As anti-VDAC antibodies detected two bands in the VDAC immunoblot, we quantified the bands separately. The total amount of VDAC in the immunoprecipitation was normalized to the targeted HK1 immunoprecipitation (which did not change; Fig 3) to yield a VDAC/HK1 ratio for each experiment. We found that a 30 min incubation with 100 µM of the N-terminal HK peptide reduced HK1-VDAC binding by 50.4 ± 5.61% (48.9 ± 6.89% in the upper band, 66.1 ± 4.27% in the lower band) in crude synaptosomes prepared from adult rat cortex (Fig. 3). The reduction with peptide treatment compared to control was significant for total VDAC as well as for each individual band. The reductions observed for each band were not significantly different from each other.

Figure 3.

An N-terminal HK peptide displaces HK1 from VDAC in synaptosomes prepared from adult rat cortex. Representative western blot of HK1 IPs under control and HK-peptide conditions labeled for HK1 and VDAC (A). The amount of HK1 found in the immunoprecipitates was not significantly affected by the HK peptide (93.6% of control, p=0.781), but normalizing VDAC immunoreactivity to HK1 immunoreactivity (VDAC/HK1 ratio) from each IP slightly reduced the variance. Therefore VDAC/HK1 in each experiment was expressed as percent of control (B). Values are displayed as mean ± SEM; n=4; ***p<0.001.

Effects of disruption of HK1-VDAC interaction on the GLT-1/mitochondria protein complex and GLT-1 function

In these same experiments that resulted in a 50% reduction of HK1-VDAC binding, we performed immunoprecipitations targeting GLT-1 and the inner-mitochondrial membrane protein, ANT, to quantify the interaction between GLT-1 and mitochondria. ANT was previously found to co-immunoprecipitate with GLT-1 (Genda et al. 2011), and also interacts with VDAC (Vyssokikh et al. 2001). In GLT-1 immunoprecipitations, co-immunoprecipitation of ANT or the matrix protein UQCRC2 was not affected by incubation with the HK peptide (Fig. 4A). Similarly, in immunoprecipitations targeting ANT, co-immunoprecipitation of GLT-1 was unaltered by HK peptide incubation (Fig. 4B). These data indicate that displacing 50% of HK1 from VDAC is not sufficient to alter the interaction between GLT-1 and mitochondria.

Glutamate uptake in astrocytes is intimately linked to both glycolytic and oxidative metabolism (Fernandez-Moncada and Barros 2014; Genda et al. 2011; Loaiza et al. 2003; McKenna et al. 1996; Pellerin and Magistretti 1994; Yu et al. 1982), which are in turn supported by HK-VDAC binding (BeltrandelRio and Wilson 1991; Rodrigues-Ferreira et al. 2012). Therefore, we also investigated the effect of disrupting the interaction between HK1 and VDAC on glutamate uptake in the same sets of experiments. We found that disruption of the HK1-VDAC interaction significantly reduced Na⁺-dependent L-[³H]-glutamate uptake to approximately 70% of control (Fig. 4C).

Discussion

The assembly of proteins into macromolecular complexes is thought to both increase the specificity of reactions and improve the speed of reactions that are dependent upon multiple proteins by limiting the impact of diffusion (Levitan 2006. Converging lines of evidence provide examples of macromolecular protein complexes coupling plasma membrane transport proteins with metabolic support. In erythrocytes, glycolytic enzymes including phosphoglycerate kinase (Campanella et al. 2005; Mercer and Dunham 1981; Parker and Hoffman 1967), glyceraldehyde phosphate dehydrogenase (GAPDH), and aldolase (Campanella et al. 2005) localize to the plasma membrane and physically interact with membrane proteins such as the anion exchanger 1 (AE1) (Campanella et al. 2005). In this system these proteins contribute to a membrane-associated pool that fuels the Na⁺/K⁺-ATPase (Mercer and Dunham 1981). The regulation of the glucose transporter GLUT4 by insulin is reciprocally controlled by its interactions with GAPDH or HK2 (Zaid et al. 2009). Previously, we conducted a proteomic analysis of GLT-1 with the purpose of identifying proteins that might be involved in regulating the localization and activity of this astrocytic glutamate transporter. This analysis revealed the presence of subunits of the Na⁺/K⁺-ATPase, a number of glycolytic enzymes (HK1, phosphofructokinase, glyceraldehyde dehydrogenase, etc.) and mitochondrial proteins (e.g. UQCRC2, ANT) in GLT-1 immunoprecipitates (Genda et al. 2011). Since then, these results have been extended to the other glial glutamate transporter, GLAST (Bauer et al. 2012), and to human tissue (Shan et al. 2014). These results suggest that the protein interactions observed in astrocytes might serve to fuel glutamate uptake and the concomitant Na⁺/K⁺-ATPase.

Based on these results, we posited the existence of a macromolecular assembly of proteins that might fuel glutamate transport, as illustrated in Figure 5. Several of the proteins identified have previously been shown to interact. Evidence for interaction between GLT-1 and the Na⁺/K⁺-ATPase in the plasma membrane preceded our own results (Fig. 5, arrow #1) (Rose et al. 2009, Genda et al. 2011). Hexokinase is bound to the outer mitochondrial membrane in part by its association with the outer membrane resident protein, VDAC (Fig. 5, arrow #2) (Sul and Wilson 1997, Abu-Hamad 2008). VDAC in turn physically interacts with the inner mitochondrial membrane protein ANT (Fig. 5, arrow #3) (Beutner 1998), Crompton 1998). Based on these known interactions, we proposed that the cytosolic enzyme hexokinase might serve to physically coordinate the interaction of the plasma membrane-bound transporter with the outer mitochondrial membrane protein VDAC and through this to the inner mitochondrial membrane protein, ANT, and the matrix protein, UQCRC2. Here, we have extended this work by validating the interaction between GLT-1, HK1, and VDAC, and we also tested the hypothesis that the interaction between HK1 and VDAC facilitates GLT-1 binding to mitochondrial proteins.

Figure 5.

Schematic depiction of the interaction between GLT-1, HK1, and mitochondrial proteins. Arrows (#1–3) refer to know protein interactions. A potential model whereby GLT1-1 is linked to mitochondrial proteins is diagramed (arrow #4). Figure is adapted from Genda et al 2011.

We have considered the possibility that the co-immunoprecipitations between GLT-1, glycolytic enzymes, and mitochondrial proteins could be an artifact due to post-solubilization aggregation. However, as shown in Figure 2, HK1 and GLT-1 co-localize within astrocytic processes in cortex. Furthermore, we previously demonstrated that neuronal activity significantly increases the probability that GLT-1 and mitochondria will be apposed in perisynaptic astrocytic processes (Jackson et al. 2014). Similarly, co-culturing astrocytes with neurons increases the overlap of GLT-1 with mitochondria in culture (Ugbode et al. 2014). The close proximity between GLT-1, HK1, and mitochondria, and its regulation by neuronal activity, provide evidence for an in vivo interaction between these proteins.

To test whether HK1 serves as a cytosolic link between GLT-1 and mitochondria, we measured the extent of GLT-1 co-immunoprecipitation with mitochondrial proteins while displacing HK1 from the outer mitochondrial membrane protein VDAC. Displacing 50% of HK1 from VDAC did not alter the interaction between GLT-1 and mitochondrial proteins (ANT and UQCRC2; Fig. 4A,B). There are two possible explanations for these results: either HK1 does not serve as a cytosolic bridge between GLT-1 and mitochondrial proteins, or the pool of HK1 that serves as the bridge possesses a higher affinity for VDAC. Displacement of 50% of HK1 from VDAC may not be sufficient to abrogate this interaction. Unfortunately, this possibility could not be tested experimentally, as 100 µM of peptide approaches the solubility limit of the peptide in our buffer. However, we favor the hypothesis that HK1 does not serve as a cytosolic bridge between GLT-1 and mitochondrial proteins. Rather, HK1/GLT-1 interactions are likely indirect through mutual interaction with VDAC. HK1/GLT-1 interactions independent of HK1/VDAC interactions are less likely, as we have previously shown that exogenous HK1 is completely relegated to mitochondria in astrocytes (Genda et al. 2011). GLT-1/VDAC interaction could be direct, or through a connecting cytosolic protein or protein complex (Fig 5, arrow #4). Lastly, it is possible that GLT-1 may interact with these proteins through its interaction with the Na⁺/K⁺-ATPase.

The glutamate transporters couple the inward movement of glutamate to the Na⁺ and K⁺ electrochemical gradients. Specifically, the inward movement of one molecule of glutamate is accompanied by the inward movement of 3 Na⁺ and a H⁺ and the counter-transport of a K⁺ ion (Zerangue and Kavanaugh 1996). The Na⁺ and K⁺ gradients that enable this exchange are established, in part, by the actions of the Na⁺/K⁺-ATPase, which moves 3 Na⁺ ions into the cellin exchange for 2 K⁺ ions for every ATP hydrolyzed. These molecules function interdependently, as inhibitors of the Na⁺/K⁺-ATPase inhibit glutamate uptake (Pellerin and Magistretti 1994) and glutamate uptake stimulates Na⁺/K⁺-ATPase activity (Pellerin and Magistretti 1997). Acute energy depletion results in a reversal of the glutamate transporters, resulting in an increase in extracellular glutamate; an effect attributed to changes in the Na⁺ and K⁺ electrochemical gradients (Jabaudon et al. 2000; Rossi et al. 2000). This functional interdependence is further supported by the recent demonstration that the glial glutamate transporters physically interact with members of the Na⁺/K⁺-ATPase family (Bauer et al. 2012; Genda et al. 2011; Illarionava et al. 2014; Rose et al. 2009).

GLT-1 alone accounts for approximately 1% of brain protein (Lehre and Danbolt 1998). The dependence of GLT-1 upon the Na⁺ electrochemical gradient, and its functional and physical coupling to the Na⁺/K⁺-ATPase, is assumed to impose a metabolic demand upon astrocytes. Both glycolysis and oxidative phosphorylation can fuel gutamate uptake and the Na⁺/K⁺-ATPase (Fernandez-Moncada and Barros 2014; Genda et al. 2011). Conversely, glutamate uptake and Na⁺/K⁺-ATPase activity stimulate glucose uptake (Hamai et al. 1999; Loaiza et al. 2003), glycolysis (Pellerin and Magistretti 1994), and oxidative ATP production (Magi et al. 2013). As an alternative to utilizing glucose, glutamate may be converted to alpha-ketoglutarate by glutamate dehydrogenase and subsequently oxidized within the mitochondria (McKenna et al. 1996; Yu et al. 1982). Increasing extracellular glutamate increases the fraction of glutamate that is oxidized in astrocytes (McKenna et al. 1996; Yu et al. 1982), and inhibitors of glutamate dehydrogenase block glutamate uptake (Whitelaw and Robinson 2013). These results indicate that a portion of transported glutamate is oxidized in astrocytes, generating energy that could be used to fuel its own uptake (Dienel 2013); Dienel and McKenna 2014). In astrocytes, it appears that glutamate uptake is deeply integrated with metabolism.

Displacing HK1 from VDAC alters glutamate uptake without affecting protein interactions between GLT - 1 and mitochondria. Disrupting HK1/VDAC binding changes ANT to a confirmation that reduces ATP/ADP translocation (for review see Vyssokikh and Brdiczka 2003), which contributes to impairment of both glycolytic and oxidative metabolism (BeltrandelRio and Wilson 1991; Rodrigues-Ferreira et al. 2012). Knockout or inhibition of ANT reduces glutamate uptake, and increased ANT expression in reactive astrocytes is associated with increased glutamate uptake (Buck et al. 2003). Glutamate uptake can be supported by either glycolytic or oxidative ATP production, but inhibiting both reduces glutamate uptake (Genda et al. 2011). Thus, the metabolic consequences of displacing HK1 from VDAC are likely responsible for the impaired glutamate uptake we observed.

Numerous interactions link the uptake and metabolism of glutamate to that of glucose. Glutamate uptake and Na⁺/K⁺-ATPase activity are fueled by both glycolysis and oxidative phosphorylation (Fernandez-Moncada and Barros 2014; Genda et al. 2011). Glutamate uptake stimulates glutamate oxidation (Sonnewald and McKenna 2002; Yu et al. 1982), glycolysis (Pellerin and Magistretti 1994), glucose uptake (Hamai et al. 1999; Loaiza et al. 2003), and glycogen formation (Hamai et al. 1999; Swanson et al. 1990). L-glutamate itself can be incorporated into glycogen (Schmoll et al. 1995). Since hexokinase catalyzes the first step of glucose metabolism (i.e. glycolysis, oxidative phosphorylation, pentose phosphate pathway, or glycogen formation), physically linking GLT-1 and HK1 may represent a mechanism to coordinate glutamate and glucose metabolism. Perhaps the involvement of HK1 in a GLT-1/mitochondria complex facilitates glutamate oxidation, and thus favors the incorporation of glucose-6-phosphate into glycogen in astrocytes by reducing the demand for glucose-derived pyruvate.

In summary, we provide evidence that HK1 and VDAC interact with GLT-1. The interaction of GLT-1 with mitochondrial proteins does not depend upon the interaction of HK1 with VDAC. However, the interaction of HK1 with VDAC supports glutamate uptake. These results highlight the close integration of glutamate uptake and astrocyte metabolism.