GLT-1 Promoter Activity in Astrocytes and Neurons of Mouse Hippocampus and Somatic Sensory Cortex

GLT-1 eGFP BAC reporter transgenic adult mice were used to detect GLT-1 gene expression in individual cells of CA1, CA3 and SI, and eGFP fluorescence was measured to analyze quantitatively GLT-1 promoter activity in different cells of neocortex and hippocampus. Virtually all GFAP+ astrocytes were eGFP+; we also found that about 80% of neurons in CA3 pyramidal layer, 10–70% of neurons in I-VI layers of SI and rare neurons in all strata of CA1 and in strata oriens and radiatum of CA3 were eGFP+. Analysis of eGFP intensity showed that astrocytes had a higher GLT-1 promoter activity in SI than in CA1 and CA3, and that neurons had the highest levels of GLT-1 promoter activity in CA3 stratum pyramidale and in layer VI of SI. Finally, we observed that the intensity of GLT-1 promoter activity in neurons is 1–20% of that measured in astrocytes. These results showed that in the hippocampus and neocortex GLT-1 promoter activity is observed in astrocytes and neurons, detailed the distribution of GLT-1 expressing neurons, and indicated that GLT-1 promoter activity in both astrocytes and neurons varies in different brain regions.


DATA COLLECTION
Data were collected from CA1, CA3 and fi rst somatic sensory cortex (SI) using a Leica (TCS SP2) confocal laser microscope. To improve signal/noise ratio, 10 frames of each image were averaged. For low magnifi cation confocal microscopy, images were collected with an air dry 10× lens and aligned to obtain composite images of hippocampus and SI using Adobe Photoshop CS2 (Adobe System, San Jose, CA, USA). For high resolution confocal microscopy (applied to eGFP co-localization and intensity studies; see below) images were acquired with a HCX PL APO 63× oil-immersion lens (1.4 numerical aperture), a pinhole of 1 Airy Unit, an image size of 512 × 512 pixels and a pixel size of 0.423 µm, with the exception of the astrocyte intensity studies in which the pixel size was 0.230 µm. Acquisition of astrocytic and neuronal eGFP signals was optimized through the 'Q LUT' button, which permitted direct visualization of pixel saturation; photomultiplier gain was set so that the brightest pixels were just slightly below saturation, and the offset such that the darkest pixels were just above zero. To avoid bleed-trough between green and red fl uorescence, images were acquired sequentially.

eGFP co-localization studies
For astrocytes, eGFP/GFAP fi elds were randomly selected from stratum oriens (so), pyramidale (sp), and radiatum (sr) of CA1 and CA3 (6 fi elds/layer/animal for each region). In SI, microscopic fi elds (10 fi elds/layer/animal) were randomly selected from layer I and VI, where GFAP immunoreactivity is robust (Hajós, 2008). For neurons, sampling of eGFP/NeuN fi elds of CA1 and CA3 was carried out as for eGFP/GFAP studies, whereas in SI microscopic fi elds were sampled in all cortical layers (10 fi elds/layer/animal).

eGFP intensity studies
For the analysis of eGFP intensity in both astrocytes and neurons we used eGFP/NeuN labelled sections. Sampling was carried out in all hippocampal and cortical layers (10 fi elds/layer/animal for each hippocampal region and 10 fi elds/layer/animal for SI). We used eGFP positive (+)-NeuN+ cells for the analysis of eGFP intensity in neurons, and eGFP+-NeuN-cells for that of astrocytes. The rationale for considering eGFP+-NeuN-cells as astrocytes is that in pilot studies we observed that in layers I-VI of SI and in strata oriens, pyramidale, and radiatum of CA1 and CA3 CNPase+ cells (i.e., oligodendrocytes) and CD11b+ cells (i.e., microglial cells) were never eGFP+. Details on immunocytochemical procedure are given in the legend to Figure 2.

eGFP co-localization studies
For each merged image, a threshold level of fl uorescence for green and red channel was determined by setting a background intensity value. Background intensity values for green and red channel corresponded to 0.5 of the median of single channel intensity (estimated with Adobe Photoshop CS2); then, background values were subtracted from each channel, yielding a thresholded image. Background intensity was chosen within a range of values (from 0.25 to 0.75 of the median) which did not affect measurement accuracy (Melone et al., 2005;Ronneberger et al., 2008); background subtraction resulted in null intensity pixel value in eGFP-structures, e.g. blood vessels (Regan et al., 2007).
Thresholded images were split into green and red channel and then analyzed with NIH ImageJ Software (http://rsb.info.nih. gov/ij). A region of interest (ROI) of 5 × 5 µm was used to asses pixel positivity within neuron somata on the green channel, while a ROI of 3 × 3 µm was used for astrocytes. Cells with an average pixel intensity ≥1 arbitrary units were considered positive for eGFP expression and were therefore included in the analysis. Colocalization of eGFP+ cells with NeuN and GFAP was calculated as the percentage of eGFP+/NeuN+ and eGFP+/GFAP+ cells on the total of NeuN+ and GFAP+ cells.

eGFP cellular intensity studies
In these studies, as in eGFP co-localization studies, background value was set at 0.5 of the median level of the entire image. In a series of pilot studies, we studied the relationship between eGFP intensity values of cells and mean intensity of the whole microscopic fi eld; we found that astrocytic and neuronal intensity levels measured were positively correlated to the mean intensity levels of acquired images (r = 0.965 for astrocytes, P = 0.008; r = 0.860 for neurons, P = 0.028; n = 10 images). Therefore, in order to obtain intensity values not affected by the mean intensity of images, we transformed intensity measures by matching the cellular with the background intensity (I cell /I background ), as described by Marrs et al. (2001). After this transformation, correlation between intensity values was not signifi cant (P = 0.226 for astrocytes; P = 0.103 for neurons), thus supporting the validity of the procedure. Then we normalized the data using the formula: where v stood for the value of I cell /I bacground to be normalized, NI v for Normalized Intensity of v, v min and v max for minimum and maximum values of v, NI min and NI max are the new minimum and maximum normalized values of v, i.e. 0 and 1 (Han and Kamber, 2006).

STATISTICAL ANALYSIS
Comparison between CA1, CA3 and SI (obtained by collapsing data from all layers of each region) was performed by one way ANOVA (α = 0.05) with Bonferroni-Dunn post test (α = 0.017 due to Bonferroni correction for multiple comparisons). Comparisons among layers within CA1 and CA3 were performed with one way ANOVA (α = 0.05) with Bonferroni-Dunn post test (α = 0.017).
Comparison between CA1 and CA3 data was assessed also with two way ANOVA considering hippocampal regions as the between factor and hippocampal layers as the within factor. Post tests used to study signifi cant differences among levels of the between and the within factors were paired t-test and one way ANOVA. Comparison among layers of SI was assessed by one way repeated measures ANOVA (α = 0.05) and Bonferroni-Dunn post test (α = 0.003) was applied. Statistical analysis was performed with SPSS (v.13.0; SPSS Inc, IL, USA).

RESULTS
In line with the description of Regan et al. (2007), we observed numerous cells displaying GLT-1 promoter activity (eGFP+ cells) in SI and hippocampus (Figures 1A,B). Double-labelling studies with GFAP and NeuN antibodies (see below for details) showed that eGFP+ cells were in most cases astrocytes and not rarely neurons ( Figures 1C,D), whereas double-labelling studies with CNPase and CD11b antibodies revealed that in SI and hippocampal gray matter eGFP+ cells were never oligodendrocytes or microglial cells (Figure 2). In addition, we observed that the intensity of eGFP+ cells varied remarkably, with neurons exhibiting low eGFP fl uorescence intensity (Figures 1E,F). We therefore studied quantitatively both the presence and the intensity of eGFP fl uorescence in astrocytes and neurons of CA1, CA3 and SI of adult GLT-1-eGFP BAC reporter mice.

INTENSITY OF eGFP-GLT-1 FLUORESCENCE IN ASTROCYTES AND NEURONS
Since astrocytes and neurons displayed different levels of eGFP fl uorescence intensity, i.e., of GLT-1 promoter activity (Figures 1E,F), we analyzed eGFP fl uorescence intensity in different layers of CA1, CA3, and SI (Table 1). To do this, we measured fl uorescence intensity within eGFP+ cell bodies and calculated its normalized intensity (NI), as described in Material and Methods. Mean NI of astrocytes was different between regions (P < 0.0001): it was higher in SI than in CA1 (∼35%, P < 0.0001) and CA3 (∼47%, P < 0.0001), but it was similar in CA1 and CA3 (P = 0.13) (Figure 5A). Mean NI of neurons was different between regions (P = 0.0492) and higher in SI than in CA1 (up to ∼50%, P < 0.0146) but not in CA3 ( Figure 5B).

DISCUSSION
These studies in GLT-1 eGFP BAC reporter transgenic mice showed that virtually all GFAP+ astrocytes and numerous neurons exhibited GLT-1 promoter activity; that the degree of activation was much higher in astrocytes than in neurons; and that GLT-1 promoter activity was highly variable across regions and layers.

GLT-1 PROMOTER ACTIVITY IN ASTROCYTES AND NEURONS
There is a long history, albeit a controversial one, on the cellular localization of GLT-1, the major glutamate transporter in the mammalian brain, and particularly on its postulated neuronal expres-sion, as briefl y summarized in Introduction. Here, we have used GLT-1 eGFP BAC reporter transgenic mice to investigate GLT-1 promoter activation in astrocytes and neurons, as this preparation allows both the visualization of individual cells whose GLT-1 promoter is active and its quantitation. The latter feature appears particularly attractive on considering that GLT-1 eGFP fl uorescence appears well correlated to GLT-1 mRNA (Yang et al., 2009), GLT-1 protein expression (Regan et al., 2007), and, most importantly, to GLT-1 function (Regan et al., 2007).
We showed that virtually all GFAP+ astrocytes expressed GLT-1 promoter activity, in line with existing data on GLT-1 localization in hippocampus and neocortex (Rothstein et al., 1994;Chaudhry et al., 1995;Lehre et al., 1995;Schmitt et al., 1996;Danbolt, 2001), and confi rmed that astrocytes are the predominant cell type expressing GLT-1. We also showed that intensity of eGFP-GLT-1 fl uorescence in astrocytes was much higher than in neurons; in CA3 sp and in layer VI of SI fl uorescence intensity in neurons represented 4% and 7% of astrocyte values, respectively. Interestingly, Furness et al. (2008) have recently estimated that GLT-1 protein levels in axon terminals are about 10% of those measured in astrocytes. The similarity between our data and those of Furness et al. (2008) provides further support to the notion that the degree of GLT-1 promoter activity appears correlated to protein expression (Regan et al., 2007).
We have also observed that a considerable number of neurons in CA3 pyramidal layer (∼ 80%) and in layers I-VI of SI (10-70%) exhibits GLT-1 promoter activity. These results extend those of previous studies that described qualitatively the presence of GTL-1 mRNA FIGURE 2 | In layers I-VI of SI and in strata oriens, pyramidale and radiatum of CA1 and CA3 cells immunoreactive to CNPase (A; n = 180), a marker of oligodendrocytes, or CD11b (B; n = 130), a marker of microglial cells, were not eGFP+. Sections were incubated in NGS (10% in PB; 1 hr) with 0.2% Triton-X and then overnight at room temperature in a solution containing primary antibodies directed to CNPase (1:100; Millipore; two mice, two sections/animal) or CD11b (1:100; Serotec; two mice, two sections/animal). Anti-mouse TRITC-conjugated secondary antibody (1 h; 1:150; Molecular Probes) were used. In control experiments primary antibodies were omitted. (A) SI layer VI; (B) CA1 stratum oriens. Scale bars: 10 µm.

FIGURE 3 | Percentage of NeuN/eGFP co-localization varies across different cortical regions.
NeuN/eGFP+ double-labelled cells are signifi cantly more numerous in CA3 and SI than in CA1. Values are mean ± sem. Data collapsed from all layers of each region from six animals. ***P < 0.001. FIGURE 5 | Regional analysis of fl uorescence intensity in astrocytes and neurons. (A) shows normalized intensity (NI) in astrocytes from CA1, CA3, and SI; (B) illustrates NI in neurons from the same regions. Note that neuronal NI is lower than astrocytic NI in all regions. Values are mean ± sem. ***P < 0.001.
in hippocampal and neocortical neurons (Torp et al., 1994(Torp et al., , 1997Schmitt et al., 1996;Berger and Hediger, 1998;Chen et al., 2004;Berger et al., 2005), and are compatible with the unexpected observation that ∼ ¾ of nerve terminals in CA1 stratum radiatum (Schaffer collaterals from CA3 pyramidal neurons) exhibit GLT-1-mediated d-aspartate uptake (Furness et al., 2008). Therefore, it is conceivable that the activation of eGFP GLT-1 promoter observed in more than 80% of pyramidal neurons of CA3 may be translated into GLT-1 protein in nerve terminals forming synapses in CA1, although this remains to be demonstrated. The discrepancy between the distribution of GLT-1 promoter activity revelead by eGFP positivity and that of GLT-1 protein reported in numerous immunocytochemical studies is conceivably related to the low levels of GLT-1 protein found in neurons, which makes it diffi cult to detect it immunocytochemically (Schmitt et al., 1996;Chen et al., 2004;Furness et al., 2008). This view is supported by the observation that eGFP-GLT-1 fl uorescence in neurons was in general very low (present results) and that an high resolution confocal microscopy analysis of terminals expressing vesicular glutamate transporter 1 (VGLUT1) in CA1 stratum radiatum showed that only few of them (<5%) were also GLT-1 positive (de Vivo, Melone, Rothstein, and Conti unpublished observations). Alternatively, it is possible that in neurons exhibiting low levels of eGFP-GLT-1 fl uorescence the activation of the promoter is not followed by production of GLT-1 protein.
The percentage of neurons exhibiting eGFP-GLT-1 activity was signifi cantly higher in CA3 pyramidal neurons and in layer VI neurons than in those of other regions or layers. Whereas CA3 pyramids were known to express GLT-1 at their axon terminals (Chen et al., to the fi ltering of the thalamo-cortical fl ow of information, it is conceivable that GLT-1 may play a role in the gating of thalamicallymediated afferent information to the neocortex.

VARIATIONS OF eGFP-GLT-1 INTENSITY
Quantitative analysis of eGFP-GLT-1 fl uorescence intensity in hippocampal and neocortical neurons and astrocytes revealed significant variations across brain regions and layers as well as between cells of the same region or layer, as previously noticed (Rothstein et al., 1994;Lehre et al., 1995;Milton et al., 1997;Regan et al., 2007;Holmseth et al., 2009). Mean normalized fl uorescence values indicated that astrocytes in SI had a higher GLT-1 promoter activity than CA1 and CA3 astrocytes (SI > CA1 > CA3). Moreover, laminar analysis of astrocytes eGFP promoter activation in different SI layers showed a similar degree of activation, whereas in CA1 and CA3 signifi cant differences were observed between strata oriens, pyramidale and radiatum. Neurons had the highest level of eGFP fl uorescence in layer VI of SI and in CA3 stratum pyramidale, and showed laminar differences only in SI.
A large number of endogenous and exogenous molecules regulates GLT-1 expression in in vitro and in vivo conditions (Gegelashvili and Schousboe, 1997;Robinson, 1998;Anderson and Swanson, 2000;Beart and O'Shea, 2007;Lauriat and McInnes, 2007). In human astrocytes several GLT-1 regulators, including EGF, TGF-α, and dBcAMP, exert their effects by acting on the GLT-1 promoter (Su et al., 2003;Li et al., 2006). If the same transcriptional mechanism operates in neurons, the high variability of eGFP GLT-1 observed here in astrocytes and neurons refl ects in all likelihood the differential regulation of GLT-1 promoter. It is conceivable that such a differential regulation, though largely constitutive, is at least in part induced dynamically by differential homeostatic needs in both physiological and pathophysiological conditions, in line with the crucial role GLT-1 plays in regulating extracellular glutamate and in shaping excitatory transmission (Tong and Jahr, 1994;Conti and Weinberg, 1999;Danbolt, 2001;Tzingounis and Wadiche, 2007). In this context, it is worth mentioning that a few dozen transporters distributed quasi-randomly inside the cleft attenuate AMPARs activation (Zheng et al., 2008), and that GLT-1 up-regulation affects synaptic plasticity (Omrani et al., 2009).
Thus, the variability of eGFP intensity described here raises the possibility that GLT-1 eGFP BAC reporter transgenic mice may reveal themselves useful tools to investigate quantitatively the regulation of GLT-1 promoter activation in in vivo physiological or pathophysiological conditions.

ACKNOWLEDGMENTS
Supported by MIUR and Università Politecnica delle Marche.