NEURONAL ACTIVITY REGULATES GLUTAMATE TRANSPORTER DYNAMICS IN DEVELOPING ASTROCYTES

Organotypic tissue slice cultures

Organotypic hippocampal slice cultures were prepared as described previously (Stoppini et al., 1991; Benediktsson et al., 2005). Briefly, postnatal day (P) 5 to P7 Sprague Dawley rat pups were killed by swift decapitation, in accordance with institutional guidelines and as approved by the animal care and use committee. Brains were placed in cold (4°C) Hanks balanced salt solution (HBSS; Gibco) supplemented with dextrose (6 mg/ml). Hemispheres were separated and hippocampi were isolated and cut into 300 μm thick transverse sections using a manual tissue chopper (Stoelting). Slices were placed onto Falcon polyethylene terephthalate-etched membrane culture inserts perforated with 1-μm pores (BD Biosciences). Slices were maintained in filter culture media (FCM) containing: 25% HBSS, 50% MEM, 25% heat-inactivated horse serum, 2mM glutamine, 0.044% NaHCO₃ and 10 U/ml penicillin-streptomycin. Tissues were cultured in a 5% CO₂ humidified incubator (36°C) for 2-14 days before use. Media was exchanged every 2 d.

Immunohistochemistry

Immunohistochemistry (IHC) was performed as described previously (Benediktsson et al., 2005). Briefly, acutely isolated tissue slices (300 μm thick), or cultured slices on membrane inserts, were placed in warm culture media containing 4% paraformaldehyde for 15 min at room temperature. Tissues were rinsed in 0.01 M PBS, extracted overnight with 1% Triton-X, blocked sequentially in 50 mM NH₄Cl and 20% horse serum (30 min each), rinsed with PBS containing 1% horse serum (PBS+HS), and incubated for 10-16 hr (4°C) in primary antibodies at the indicated dilutions: anti-GFAP mouse monoclonal (Sigma, catalog #G3893), 1:500; anti-GLT-1 rabbit polyclonal (Rothstein et al., 1994), 1:500; anti-GLT-1 guinea pig polyclonal (Chemicon, AB1783), 1:10,000; anti-synapsin-1 rabbit polyclonal (Invitrogen), 1:200; anti-GFP mouse monoclonal (Sigma, G6539), 1:500. Tissue slices were rinsed (3 × 15 min) with PBS+HS, then incubated 10-16 hours (4°C) in Alexa-Fluor-488 or -568 conjugated secondary antibodies (Invitrogen), diluted 1:300 in PBS+HS, before final rinses and mounting.

Plasmid construction and characterization of GLT-1-GFP

Two different plasmids were used: a red fluorescent protein (DsRed2; Clontech) and a full length GLT-1 fused to GFP (GLT-1-GFP). To construct GFP-GLT-1, GFP was fused to the N-terminus of GLT-1 by inserting it in frame into the pEGFP-C3 vector (Clontech) between SalI and BamH1 restriction sites. For both plasmids, transcription was driven by a cytomegalovirus (CMV) promoter. Plasmids for transfection were purified using a Plasmid Maxi Kit (Qiagen). Optimal DNA concentrations for gene gun bullets were between 0.9-1.5 μg/μl. The ability of the GLT-1-GFP fusion protein to support glutamate uptake was evaluated after transient transfection of plasmid into HEK cells, as described (Davis et al., 1998). Cells were incubated with 1, 5, 10, 50 or 500 μM ³[H]-L-glutamate at 37° C for 4 min, washed with PBS, dissolved in 0.1 N NaOH, and the radioactivity present in a 500 μl sample was measured with a scintillation counter. GLT-1-GFP supported saturable, Na⁺-dependent uptake of glutamate with a K_m = 35 μM, comparable to that exhibited by wild type GLT-1 (Tan et al., 1999).

Preparation of cDNA carriers and biolistic transfection in hippocampal tissue slices

Biolistic transfection was performed using a Helios gene gun (Bio-Rad) as described in detail previously (Marrs et al., 2001; Benediktsson et al., 2005). Colloidal gold particles (1.0 μm diameter) were used as the plasmid carrier. DNA was added to gold to obtain the optimal ratio of DNA:gold, referred to as the DNA loading ratio [DLR = (concentration of DNA × volume used) / weight of gold]. Typically a DLR of 2.0 was used, although somewhat higher transfection rates and expression levels could be achieved with a DLR up to 6.0. To make double bullets containing two different plasmids, the optimal DLR ratio was 1:3 for GFP-GLT-1: DsRed2. Gene gun “bullets” were prepared using a Bio-Rad tubing prep station. For biolistic transfection, the gene gun was shot from a distance of 1-2 cm at a pressure of 80-100 psi. Cultures were transfected 1-12 days after slicing and returned to the CO₂ incubator. Typically, slices were imaged 24-48 hr after transfection.

Treatment of organotypic tissue slices with toxins

Slice cultures were exposed to one of 3 conditions: (1) control slices were cultured in FCM alone for 7 d ( “CTL”); (2) slices were cultured for 7 d in 0.2 μM tetrodotoxin (TTX; RBI) diluted in FCM (FCM+TTX); (3) slices were cultured for 6 d in FCM, then switched to 10 μM SR-95531 (gabazine, GBZ; Sigma) in FCM (FCM+GBZ) for 24 hr prior to imaging. All media was exchanged every 48 hr. Although GBZ treatment induces persisting patterns of enhanced network activity in slice cultures, this does not cause detectable cell death or other obvious features of pathology (Zha et al., 2005) presumably because developing hippocampal tissues are resistant to seizure-induced cell death (Haas et al., 2001).

Confocal imaging and image processing

Images were captured using a Leica TCS NT scanning laser confocal microscope (inverted) equipped with Argon (Ar; 488 nm), Krypton (Kr; 568 nm), and Helium-Neon (He-Ne; 633 nm) lasers, or a Leica SP2 AOBS confocal system equipped with Ar, green He-Ne (543 nm), orange He-Ne (594 nm), and red He-Ne (633 nm) lasers. GFP-fusion proteins were visualized using the 488 nm Ar laser line and a fluorescein-like filter set, and DsRed2 was imaged using Ar or Kr laser excitation and a rhodamine-like filter set.

For 3-D reconstruction of cells in fixed tissue slices, stacks of 20-80 optical sections were taken (0.4-1.0 μm z-step intervals) using a 20x/0.7 Plan Apo or 63x/1.2 water Plan Apo lens. Two-dimensional (2D) projection images were generated in Image (Scion Corp.) or ImageJ (Wayne Rasband, NIH) using a maximum brightness operation.

Data analysis

To assess localization of synapsin-1 with either GLT-1 immunopuncta or GFP-GLT-1, confocal image stacks representing 10-40 μm deep tissue volumes were displayed as 2D projection images using the Leica Confocal Software. Then 2 μm circular regions of interest (ROIs) were drawn around randomly selected GLT-1 clusters (blind to the synapsin-1 channel). After a cluster was selected, intensity values within the ROIs were calculated at each focal plane through the stack for both fluorescent channels, and peak intensities of both GLT-1 clusters and synapsin-1 immunopuncta were plotted through the z-stack of images. The GLT-1 cluster was scored as “apposed” if the peak intensity of the synapsin-1 channel was within 1 μm on either side of the peak intensity of the GLT-1 channel (in the z dimension). For GFP-GLT-1 an equal number of 2 μm diameter circles were randomly placed along regions of astrocytic processes where there were no obvious GLT-1 clusters. These ROIs were analyzed for apposition in three dimensions, as above. The statistical significance between ‘off-cluster’ and ‘on-cluster’ apposition to synapses was calculated using a Student T-test. The statistical significance between treatment groups was first calculated using ANOVA, followed by a post hoc T-test.

To define GLT-1 cluster density, individual z-sections were thresholded to determine 2-D cell surface area, then the initial threshold level was doubled to capture bright clusters but not background fluorescence. Particle analysis was performed with a minimum of 1 pixel. By defining clusters as fluorescent regions at least 2× greater than the intensity of the diffuse membrane fluorescence, our analysis normalizes for any differences in expression levels between cells or treatment conditions. Where large suprathreshold areas appeared to be composed of overlapping clusters, a manual count was performed, comparing the thresholded and raw images.

All analyses were performed blind to condition. ANOVAs were calculated for each data set and then Post-hoc Student-T-tests were performed for individual sets (TTX, GBZ, and CTL).

Time-lapse imaging

For time-lapse recordings, slice cultures were mounted in a closed chamber containing HEPES-buffered culture media (500 μl), as described previously (Benediktsson et al., 2005; Dailey et al., 2011). The chamber was maintained at 35°C by warm, forced air, and media was exchanged every 2-3 hours. The chamber medium lacked phenol red and contained antioxidant (25mM ascorbate) to scavange reactive oxygen species that may contribute to photodynamic damage (Dailey et al., 2006). To capture changes in astrocyte structure, stacks of 6-20 optical sections (512 × 512 pixel array) were collected at 1-6 min intervals. In most cases, image stacks were collected at 4-6 min intervals for several hours. In some brief (<1 hr) time-lapse experiments, images were collected at 1 min intervals to capture rapid events with greater time resolution. To resolve fine details of astrocytic processes, a 63x lens was used with 1.5 – 3.5 additional electronic zoom. The z-step size between optical sections was usually ~0.8 μm (range 0.5-1.0 μm). Stacks of images were collected in the z-dimension in such a way that they were partially overlapping so that, within the volume of tissue imaged, structures do not get lost between individual images. Laser illumination was attenuated to a minimal level (10-20%) using an acousto-optical tunable filter (AOTF), and the detector pinhole was opened (two Airy units) to improve light collection. The continuously variable barrier filters on the Leica SP2 AOBS confocal were adjusted to maximize signal detection. Live cells co-expressing GFP and DsRed2 were illuminated at a single wavelength (488 nm) and detected simultaneously in two channels to reduce the light load. For improved signal-to-noise ratio, 2-3 scans were averaged at each optical section. At the beginning of each imaging session, detector gain and offset were adjusted so that the most intense fluorescence was just below detector saturation and the background (non-fluorescent) signal was just above the detection floor. Under these conditions, time-lapse recordings of healthy astrocytes were performed for 3-10 hr without evidence of photobleaching or phototoxicity, such as blebbing or changes in cell activity (Dailey et al., 2011). Projection images were filtered (Gaussian blur in Adobe Photoshop) to reduce noise, and for a given time series, all images were adjusted identically for brightness and contrast. To correct for any drift in the XY plane, projection images in time-lapse sequences were subjected to an automatic subpixel registration algorithm for rigid body motion (Thévenaz et al., 1998) using the StackReg plugin in ImageJ.

Endogenous GLT-1 exhibits a punctate distribution in developing hippocampus

Astrocytes begin to express GLT-1 prenatally (Shibata et al., 1996), and in the forebrain levels of this transporter steadily increase during the first three postnatal weeks (Furuta et al., 1997). Immunolabeling for GLT-1 has revealed dense labeling of the neuropil, particularly within synaptic layers of stratum radiatum (SR) and lacunosum moleculare of area CA1, in both developing and adult hippocampus (Furuta et al., 1997; Williams et al., 2005). We used antibodies to determine if GLT-1 is distributed similarly in developing hippocampus in vivo and in tissue slices cultured ex vivo. Anti-GLT-1 immunoreactivity (IR) showed punctate labeling in hippocampal tissues isolated from brains of developing rat (P6 and P12). During the first postnatal week (Fig. 1A-D), at low magnification GLT-1 IR was most evident as concentrated “patches” or aggregates of GLT-1 puncta above a more dispersed punctate pattern of tissue staining (Fig. 1A,B). Macroscopic patches of GLT-1 IR puncta were similar in size (60-80μm in diameter) and shape to individual protoplasmic astrocytes in developing hippocampal tissues in vitro (Benediktsson et al., 2005) and in vivo (Bushong et al., 2004). Indeed, higher magnification images of patches showed more intense GLT-1 puncta appearing over a diffuse, lower level of GLT-1 IR outlining cell bodies and branches (Fig. 1C,D). Double-labeling with both GLT-1 and GFAP antibodies confirmed that GLT-1 puncta circumscribed cell bodies and processes of astrocytes (Supplementary Fig. 1).

The patchy, astrocyte-like staining pattern was less evident as developmental age increased. In tissues harvested from second postnatal week rat (Fig. 1E,F), the punctate GLT-1 immunostaining was more homogeneous, although occasional cell-like “patches” were seen. A similar GLT-1 staining pattern was observed in tissue harvested from one-week-old rats and cultured for one week, suggesting that a developmental change in GLT-1 distribution also occurs in vitro (Fig.1G,H). The patterns and levels of GLT-1 immunoreactivity were qualitatively similar in acutely isolated (P12) and age-matched cultured tissue slices (P6 + 6 days in vitro; cf. Fig. 1E-F and G-H), suggesting that overall expression levels are similar. Previous studies have shown that synaptic density in area CA1 increases during the second postnatal week in vivo and in slice cultures (Buchs et al., 1993). Thus, the developmental change in GLT-1 distribution coincides with an increase in the density and complexity of astrocyte processes during maturation (Hama et al., 2004; also Supplementary Fig. 1), as well as with the formation and maturation of neuronal synapses.

GLT-1 clusters are increasingly localized near synapses as development proceeds

GLT-1 has been shown to limit the activation of NMDA and metabotropic glutamate receptors (Tanaka et al., 1997; Huang et al., 2004) and to alter the threshold for induction of LTP in area CA1 (Katagiri et al., 2001), suggesting that it is localized near excitatory synapses. To determine if GLT-1+ puncta were localized near synaptic structures, we co-immunostained acutely isolated hippocampal slices for GLT-1 and a presynaptic terminal protein, Synapsin-1 (syn-1; Fig. 2A,B). Three-dimensional analysis of confocal image stacks showed that at P6, 68 % of GLT-1+ clusters (n = 314) in area CA1 were within 1 μm of a syn-1+ punctum, and by P15 the degree of apposition increased to 86 % (n = 415 GLT-1 clusters) (Fig. 2C). This developmental increase in synaptic apposition of GLT-1 in vivo was recapitulated in organotypic slice cultures in vitro (Fig. 2C). Together, these results suggest that GLT-1 in astrocytes is organized into discrete clusters that are preferentially localized to sites near neural synapses.

GFP-GLT-1 forms discrete clusters in hippocampal astrocytes in situ

Excitatory synapses form rapidly between hippocampal neurons when growing axons encounter motile dendritic filopodia (Friedman et al., 2000). The expression of GLT-1 by astrocytes precedes the formation of excitatory synapses (Shibata et al., 1996), raising the possibility that preformed GLT-1 clusters could move to regions where astrocyte lamellipodia contact nascent synapses. To determine the subcellular distribution of GLT-1 in individual astrocytes and to monitor the movement of GLT-1 clusters over time, we used ballistics to randomly transfect a few astrocytes in developing organotypic slices with cDNA encoding a GFP-GLT-1 fusion protein. We restricted our analysis to astrocytes (n = 36) within the stratum radiatum of area CA1 because hippocampal slice cultures retain the intrinsic Schaffer collateral projection to area CA1 (Gahwiler et al., 1997), and this region represents a well characterized model for studying astrocyte-synapse relationships(Ventura and Harris, 1999). As expected for proteins targeted to the plasma membrane of astrocytes in situ (Benediktsson et al., 2005), expression of GFP-GLT-1 revealed the fine lamellipodia and filopodia of these complex cells (Fig. 3A). In higher magnification views of transfected astrocytes, individual GFP-GLT-1 clusters or “hotspots” of transporter accumulation were visible over a lower level, diffuse membrane fluorescence (Fig. 3B). GFP-GLT-1 clusters, defined as discrete spots of fluorescence intensity at least two times greater than the diffuse membrane fluorescence, had an average diameter of 1.1 ± 0.04 μm (n = 100). This was nearly identical to the diameter of GLT-1+ immunolabeled puncta (1.1 ± 0.16 μm; n = 40 clusters from 4 cells), indicating that expression of GFP-GLT-1 does not greatly inflate cluster size. GFP-GLT-1 clusters were observed on the cell body and along processes. GFP-GLT-1 clusters were also seen on the surfaces and edges of lamellipodia, and at the tips of filopodia and bulbous spine-like structures (Fig. 3C-F), while some fine astrocyte processes lacked GFP-GLT-1+ clusters. Analysis of individual astrocytes co-expressing GFP-GLT-1 and DsRed2, a cytosolic fluorescent protein, confirmed that GFP-GLT-1 is targeted to the plasma membrane and that GFP-GLT-1 clusters represent sites of protein concentration (Supplementary Figs. 2 and 3).

Time-lapse imaging reveals dynamic remodeling of GFP-GLT-1 clusters in developing astrocytes

To assess whether GFP-GLT-1 clusters represent stable accumulations of transporters or dynamic structures that can be remodeled over time, we examined GFP-GLT-1 clusters in astrocytes in live tissue slices using time-lapse confocal microscopy. To monitor dynamic changes in three dimensions and circumvent problems associated with focal drift, we collected stacks of confocal optical sections every 4 to 6 min for up to 10 hr. Analysis of time-lapse movies (n = 7 cells) revealed that GFP-GLT-1 clusters in developing astrocytes display a variety of dynamic behaviors (Fig. 4A). Many clusters along the shafts of astrocytic processes were mobile, moving along the edges of processes (Fig. 4B and supplementary video). Moving clusters sometimes appeared to merge or disappear from view. ‘Cumulative time’ images showed a non-uniform distribution of GLT-1 clusters over time; some regions of astrocyte branches had ‘hotspots’ with many GFP-GLT-1 clusters, whereas other regions showed a paucity of clusters (Fig. 4B Avg). At the growing tips of astrocytes, GFP-GLT-1 clusters were carried along with branches and filopodia as they extended and retracted (Fig. 4C). Although many clusters moved short distances over a period of minutes, only rarely did clusters translocate long distances along the shafts of astrocyte processes, suggesting that most clusters are integral plasma membrane components rather than intracellular transport vesicles.

When images were captured at short (1 min) time intervals, GFP-GLT-1 clusters were observed to move along with filopodia-like structures that projected laterally from astrocyte branches as they extended and retracted (Fig. 5A). Some filopodia- or spine-like structures bearing clusters at their tips were relatively stable, showing no appreciable change in length for at least several minutes (Fig. 5A,C). Although most filopodia were rather short (~2 μm), some filopodia carrying a GFP-GLT-1 cluster were over 6 μm in length (Fig. 5B,D). Such long filopodia have been observed previously on astrocytes in vivo at this developmental stage (Bushong et al., 2004), indicating that they are not abnormal structures induced by GFP-GLT-1 expression. These observations indicate that clusters of GLT-1 can move along the surface of developing astrocyte processes, and emerge with growing lamellipodial and filopodial structures.

Most GFP-GLT-1 clusters are situated adjacent to presynaptic terminals

In order to determine the relationship of GFP-GLT-1 clusters to synapses, slice cultures containing transfected astrocytes were immunostained with syn-1 antibodies. In single confocal image planes through transfected astrocytes, many but not all GFP-GLT-1 clusters were located adjacent to, or overlapping with, syn-1+ puncta (Fig. 6A,B). To further investigate this association in three-dimensions, we assessed the apposition of GFP-GLT-1 clusters and syn-1+ puncta in through-focus stacks of confocal images from 6-7 day old slice cultures. We found that 82 % of GLT-1 clusters (n = 332 clusters from 20 cells) were located within 1 μm of a syn-1+ punctum, whereas only 50 % of randomly selected regions of an astrocyte that lacked discrete GFP-GLT-1 clusters were closely adjoining syn-1+ puncta (p < 0.0001; Fig. 6C,D). These results indicate that clusters of GFP-GLT-1 are preferentially located near synapses.

Neuronal activity influences GFP-GLT-1 cluster size and density

Chronic alterations in neuronal activity are associated with changes in GluT expression (Wong et al., 2003), and transection of afferent inputs reduces GLT-1 expression in the target field (Ginsberg et al., 1995), suggesting that GluT expression can be regulated by neuronal activity. To determine if changes in neuronal activity alter GLT-1 clusters in astrocytes, we cultured organotypic slices for 7 days in either 0.2 μM tetrodotoxin (TTX), no drug (control, CTL), or for 24 hr post-transfection in 10 μM gabazine (GBZ). Examination of treated slices revealed that GFP-GLT-1 cluster density was decreased by 37% from control when action potential firing was blocked by TTX (n = 25 image planes from 6 cells, p < 0.01), whereas cluster density was increased by 25 % when neuronal activity was enhanced by GBZ (n = 36 image planes from 9 cells, p < 0.05; Fig. 7A). Furthermore, the mean size of GFP-GLT-1 clusters increased by 49 %, from 1.1 ± 0.04 μm (n = 100 clusters) to 1.6 ± 0.06 μm (n = 80 clusters) when neuronal activity was enhanced by GBZ (p < 0.0001; Fig. 7B). There was also a small but significant (p<0.01) decrease in cluster size when neural activity was blocked (TTX). These findings indicate that global changes in neuronal activity alter both the size and density of cell surface GluT clusters in developing astrocytes.

Synaptic transmission recruits endogenous GLT-1 to synaptic sites

Chronic changes in neuronal activity can induce pre- and post-synaptic homeostatic plasticity, including changes in the probability of glutamate release from nerve terminals and in the density of synapses (Turrigiano, 2007). If proper localization of GLT-1 is important for clearance of glutamate at excitatory synapses, the dynamic movement and apposition of GLT-1 clusters to synapses should also be influenced by changes in neuronal activity. Consistent with this hypothesis, there was a 53 % decrease in apposition of endogenous GLT-1 clusters with syn-1+ puncta (p < 0.01) in slices cultured in TTX (n = 262 GLT-1 clusters, 8 cells) (Fig. 8). Conversely, there was a 15 % increase (p < 0.05) in apposition between GLT-1 clusters and syn-1 puncta (n = 174 clusters, 6 cells) in slices exposed to GBZ (Fig. 8). Together, these data indicate that neural activity bi-directionally regulates the perisynaptic localization of GLT-1 clusters.

Perisynaptic localization of GLT-1 clusters

Previous EM-level analyses of rat hippocampus indicate that many spine synapses are contacted by astrocyte processes in vivo (76%, Ventura and Harris, 1999; Witcher et al., 2007) and in cultured tissue slices in vitro (85%; Lushnikova et al., 2009). Our data showing that ~82% of GLT-1 clusters are located within 1μm of a synapse are consistent with these findings, and suggest that many synaptically-associated astrocytic processes in situ contain GLT-1 clusters. It is possible that GLT-1 also may regulate glutamate levels near GABAergic synapses to alter heterosynaptic interactions. However, a previous confocal study in rat neocortex found that a larger fraction of GLT-1 is intimately associated with excitatory terminals than with inhibitory terminals immunolabeled for the GABA transporter, VGAT (Minelli et al., 2001).

Neural activity regulates GLT-1 clustering

Modification of individual synapses is thought to underlie key aspects of learning and memory, and is manifested through changes in synaptic features such as the type and number of receptors (Newpher and Ehlers, 2008), the size and shape of dendritic spines (Alvarez and Sabatini, 2007), and possibly the number of synaptic contacts (Desmond and Levy, 1986; Stewart et al., 2000). The prominent role of GLT-1 in buffering and clearing synaptic glutamate suggests that it may be co-regulated in response to changes in synaptic activity. In accordance with this hypothesis, astrocytes in vitro increase expression of GLT-1 when co-cultured with neurons (Swanson et al., 1997; Perego et al., 2000) or when exposed to neuron-conditioned media (Gegelashvili et al., 1997). Furthermore, enhancement of neuronal activity in vivo through repetitive sensory stimulation induces robust increases in GLT-1 expression in corresponding cortical regions (Genoud et al., 2006), and chronic restraint stress enhances GLT-1 expression in regions that exhibit extensive dendritic remodeling (Reagan et al., 2004). Conversely, transection of afferent fibers decreases GLT-1 levels in target zones (Ginsberg et al., 1995; Lievens et al., 2000; Hein et al., 2001), and GLT-1 expression is altered in a variety of neurological diseases, such as amyotrophic lateral sclerosis (Lin et al., 1998; Howland et al., 2002; Vanoni et al., 2004), schizophrenia (Matute et al., 2005), and following acute ischemic brain injury (Rao et al., 2001), indicating that GLT-1 expression is tightly regulated. Our results indicate that alterations in neuronal activity not only induce changes in GLT-1 expression, but also alter the spatial organization of GluTs in astrocyte membranes, pointing to an important role for synaptic activity in regulating the formation, growth, and spatial arrangement of GLT-1 clusters, including the recruitment of GLT-1 clusters to perisynaptic locations.

GLT-1and synaptic homeostasis

Changes in the expression level or position of GluTs in astrocytes may reflect homeostatic responses to neural activity. Homeosynaptic plasticity is a phenomenon whereby a global change in activity results in a “scaling” of synapse strength so that the relative strength of individual synapses can be maintained when there is a global change in activity (Turrigiano, 2007; Davis, 2006). In this model, chronic activity deprivation induces an increase in the number of synaptic glutamate receptors (GluRs), while enhanced activity causes a decrease in receptor number (Turrigiano et al., 1998; Lissin et al., 1998; O’Brien et al., 1998), an effect that can manifest at the level of individual synapses (Hou et al., 2008). We found that chronic exposure of hippocampal slices to TTX significantly reduced the size and density of GLT-1 clusters and decreased the number of these clusters that were adjacent to synapses; conversely, enhancing activity by exposure to gabazine resulted in an increase in the number of GLT-1 clusters and closer apposition of these clusters to synapses. Astrocytes therefore also exhibit homeostatic changes in response to global changes in neuronal activity, scaling back the number and density of GluTs near release sites when demand is reduced. These results suggest that the most active synapses may recruit additional GluTs to provide additional capacity for buffering or clearance, perhaps to ensure that receptors in synaptic and perisynaptic domains are not subjected to prolonged glutamate transients or spillover from adjacent, highly active synapses.

GLT-1 cluster dynamics

Our data indicate several possible modes of GLT-1 recruitment to perisynaptic sites: clusters may translocate to perisynaptic sites within stable astrocytic processes; astrocytic processes may extend towards synapses, carrying along GLT-1 clusters; or clusters may form de novo at perisynaptic sites within ensheathing astrocytic processes. The relative contributions of these different modes of delivery are presently unknown and may differ depending on the developmental and physiological context. Additional studies will be required to determine whether neural activity regulates the rates of formation, movement, or turnover of GLT-1 clusters.

In principle, localization of GLT-1 in astrocytes may involve mechanisms similar to trafficking of GluRs in neurons, in which receptors insert into the plasma membrane and shuttle between synaptic and extra-synaptic locations by lateral diffusion (see Renner et al., 2008). GluRs exhibit a range of mobility behaviors, and localization depends on interactions with cytoskeletal-associated scaffolding proteins that restrict diffusion (Newpher and Ehlers, 2008). Likewise, GLT-1 localization in astrocytes may involve the actin cytoskeleton (Zhou and Sutherland, 2004), possibly through interactions with a LIM protein, Ajuba (Marie et al., 2002).

The rates of GLT-1 clustering and remodeling are likely to have functional consequences for synaptic development and plasticity. Depending on the timing and rate of GLT-1 recruitment, nascent synapses lacking astrocytic ensheathment may have greater potential for activating extrasynaptic mGluRs and NMDA receptors, if localization of glial transporters lags behind the formation of neuronal synaptic machinery. Two previous time-lapse studies (Haber et al., 2006; Nishida and Okabe, 2007) investigating the relationship of astrocytic remodeling to neuronal spine dynamics support the notion that astrocytes can modulate the development of synaptic structures by directly contacting dendritic spine precursors (filopodia). This interaction may help stabilize newborn synapses and promote spine maturation. Our time-lapse data complement these studies by identifying mobile, preformed GluT clusters within developing astrocyte branches in situ, thereby suggesting that GluTs can be rapidly delivered to or retrieved from newly formed synapses in response to local synaptic activity.

Although the mechanisms by which synaptic transmission regulates astrocyte development and synaptic ensheathment remain to be fully elucidated, a previous study has shown that local application of glutamate induces growth of filopodia-like structures on cultured astrocytes in vitro (Cornell-Bell et al., 1990), suggesting that glutamate spillover from active synapses may induce astrocyte process outgrowth (Witcher et al., 2007). Indeed, imaging studies in tissue slices indicate that astrocytic processes near active synapses are motile, extending sheet-like and filopodia-like protrusions (Hirrlinger et al., 2004, Benediktsson et al., 2005). EM studies in slice cultures have shown that synaptic potentiation induces increased glial coverage of excitatory synapses in hippocampal area CA1 (Lushnikova et al., 2009), consistent with the hypothesis that astrocytic processes preferentially surround synapses that have more glutamate escaping from their perimeters (Ventura and Harris, 1999). Such activity-dependent remodeling of astrocyte structure and neuron-astrocyte interactions (Theodosis et al., 2008) may explain the variability in astrocyte ensheathment of excitatory synapses in the hippocampus (Ventura and Harris, 1999).

Together, our results suggest that mobile GluT clusters may contribute to rapid spatial and temporal regulation of glutamatergic signaling at the level of individual synapses. Activity-dependent recruitment and regulation of GluTs within dynamic perisynaptic glial processes supports the hypothesis that neuronal and glial components of the tripartite synapse work together to ensure optimal signaling. Greater insight into the molecular mechanisms that regulate GluT localization and remodeling may help elucidate how abnormal signaling between neuronal and glial elements contributes to developmental abnormalities and altered CNS function in disease.