Among the different types of glial cells present in the adult central nervous system (CNS), astrocytes represent the most abundantly distributed cell type throughout both white and gray matter regions. While long considered as passive supporting cells, it is now generally acknowledged that astrocytes play major roles in the control of neuronal functions. A surprising feature of astrocytes is that their functions are continuously modified throughout their life span with respect to modification of their phenotypic expression and morphological organization. Immature astrocytes in the developing CNS exhibit a bipolar organization, with elongated processes that contain bundles of intermediate filaments essentially formed of mixed polymers of nestin and vimentin (Bignami et al., 1982; Lendahl et al., 1990; Eliasson et al., 1999). A major role of these immature astrocytes (also called radial glia) is to support survival and migration of newly formed neurons (Rakic, 1972; Levitt and Rakic, 1980). During the perinatal period, changes in astroglial form occur; the bipolar radial glia disappear and are replaced by multipolar astrocytes. At this stage, the expression of the intermediate filament proteins vimentin and nestin progressively decreases, in parallel to the increased expression of another intermediate filament protein, glial fibrillary acidic protein (GFAP) (Bovolenta et al., 1984; Pixley and de Vellis, 1984; Pixley et al., 1984; Eliasson et al., 1999). Recognized roles of mature astrocytes mainly include the formation of the blood-brain barrier, maintenance of ion homeostasis, and metabolism of neurotransmitters (Fedoroff and Vernadakis, 1986; Aschner and Kimelberg, 1991). With increasing age, the astrocytic expression of GFAP continues to increase, leading to a progressive hypertrophy of astrocytes within most CNS regions (Kohama et al., 1995; David et al., 1997; Amenta et al., 1998; Legrand and Alonso, 1998; Unger, 1998). Although the consequences of such aged-related modifications of astrocytes are not clearly understood, they have been proposed to be at least partly responsible for the alteration of neuronal functions (Kullberg et al., 1988; Sykovà et al., 1998; Bacci et al., 1999).
To date, the factors that are responsible for the phenotypic and morphological modifications of astrocytes during their life span are not known. It has been extensively documented that the phenotype and the morphological organization of astrocytes can be modified by a large variety of extrinsic factors. For instance, surgical lesions to the CNS have been shown to induce a marked reaction of the astrocytes located around the lesioned area, with a marked cellular hypertrophy accompanied with the dramatic increase in the synthesis of most intermediate filament components (Eddleston and Lucke, 1993). During the past years, however, several in vivo studies have provided evidence that the morphology of astrocytes was closely related to neuronal activity (Sirevaag and Greenough, 1991; Matsutani and Leon, 1993; Canady et al., 1994; Jones and Greenough, 1996). This idea has received strong support from several studies demonstrating that astrocytes of the intact CNS express a large variety of receptors for neurotransmitters or neuropeptides (Aoki, 1992; Hösli and Hösli, 1992; Porter and McCarty, 1997).
Recently, a series of in vitro and in vivo studies has demonstrated that GABAergic signaling mediates the morphological organization of immature astrocytes and their expression of GFAP (Matsutani and Yamamoto, 1997; Mong et al., 2002). Although the distribution of astrocytes is quite homogeneous throughout the adult CNS, it is well documented that their morphological organization and their level of GFAP synthesis vary greatly from one region to the other. The present study was thus designed to determine whether differential GABAergic signaling was responsible for such differences observed between different regions of the adult rat brain. For this we first determined whether the morphological organization of astrocytes and their expression of GFAP were related with the GABA levels detected within these regions. In a second step, we evaluated astrocytic parameters under various experimental conditions aimed at increasing GABAerging signaling in vivo.
MATERIALS AND METHODS
Young adult (2- to 3-month-old) male Wistar rats (Iffa-Credo, l'Arbresle, France) were used. They were divided into three groups: untreated rats (n = 8); rats treated with either (amino-oxy)acetic acid (AOAA; n = 11), an inhibitor of GABA transaminase, or saline control (n = 9); and rats treated with either muscimol (n = 4), an agonist of GABAA receptor, or with saline control (n = 3). All the experiments were in compliance with European Communities Council Directives (86/609/EEEC).
Animals received a daily intraperitoneal (IP) injection of 0.5 ml saline either alone (controls) or containing AOAA at 10 mg/kg (AOAA-treated). In a preliminary study, two animals of each groups were killed by decapitation between 1 and 3 p.m. after 1, 4, 9, or 15 days of treatment. Three additional controls and five AOAA-treated rats were killed after 9 days of treatment. After decapitation, the brain was rapidly dissected and cut through the mid plane in two sagittal parts that were processed for either Western blot and high-performance liquid chromatography (HPLC) or for immunocytochemistry.
After deep anesthesia with equithesin (3 mg/kg), rats were placed in a stereotaxic device. A 28-gauge stainless steel cannula (Alzet brain infusion kit) was implanted into the right lateral ventricle (0.2 mm posterior to Bregma, 1.5 mm lateral to midline, and 4 mm below the surface of the skull) and connected to a 7-day Alzet osmotic minipump filled with either saline (controls, n = 3) or saline + muscimol (1 mg/ml, n = 4).
Four days after the implantation, the rats were decapitated and the brains dissected. Each brain was then rapidly cut in two rostrocaudal parts through a frontal plane passing through the cannula implantation site and were processed for either Western blot and HPLC (caudal part) or for immunocytochemistry (rostral part).
Western Blotting and HPLC Analyses
Immediately after decapitation of the animals, the brains were dissected as described above and frozen in liquid nitrogen. Less than 15 min after being frozen, the brain pieces were cut frontally into 150 μm thick sections with a Jung CM3000 cryostat. The different brain regions were microdissected according to the micropunch technique of Palkovits and Brownstein (1988) by using a cannula of 1.5 mm in diameter. The punches were transferred to tubes containing 50 μl ice-cold 0.5 M HClO4 and refrozen in liquid nitrogen. The tissue punches were stored at −80°C for up to 2 weeks. For subsequent measurements, the tubes were thawed and homogenized with a plastic eppendorf piston connected to a mechanic drill at 1,000 rpm (10 up-and-down strokes). Proteins were quantified by the method of Lowry et al. (1951). The protein amount in each punch from different regions was found to be 68.4 ± 7.2 μg (± SEM; n = 35).
For sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), homogenates containing 5 μg of protein was diluted with loading buffer and loaded into each pocket of 15 mm thick polyacrylamide SDS minigels containing a 4% polyacrylamide stacking gel. The separating gel used was at 10% (acrylamide/bisacrylamide 19:1, both for stacking and separating gels). Proteins were separated at constant ampere (30 mAmps) until the bromphenol blue front reached the end of the separating gel, for approximately 3–3.5 h. The migration was visualized by using prestained broad-range protein markers with apparent molecular weights between 6.5 and 175 kDa (New England BioLabs).
The proteins were subsequently transferred to nitrocellulose membranes (Hybond ECL, Amersham Pharmacia) by tank blotting at 80 V for 2 h. The membranes were blocked with 5% fat-free milk for 2 × 15 min. Antibody incubation was performed overnight at 4°C with either a 1:10,000 dilution of a mouse IgG monoclonal antibody against glial fibrillary acidic protein (GFAP; Sigma), followed by a 1-h incubation at room temperature with a 1:4,000 dilution of a peroxidase conjugated antimouse IgG (Sigma), or an overnight incubation with a 1:1,000 dilution of a mouse IgG monoclonal antibody against anti–glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Chemicon), followed by a 1-h incubation with a 1:4,000 dilution of a peroxidase-conjugated antimouse IgG (Sigma). ECL Western blotting reagents (Amersham Pharmacia) were used for detection of the protein recognized by the antisera. The molecular weight of proteins was visualized by creating an overlay of the membranes with blotted colored protein markers and immunoreactive signal on the film. The optical density of immunoreactive signals was estimated by using the NIH image software.
To determine the levels of GABA, 10 μl of each tissue punch homogenate was transferred to 190 μl ice-cold 0.5 M HClO4 and incubated overnight at 4°C. The tubes were centrifuged at 10,000 g for 10 min and 100 μl of the protein-free supernatant was analyzed by reverse-phase HPLC using a Waters Spherisorb ODS column (4.6 mm × 25 cm) and an electrochemical detector (Waters 460) with the working electrode potential set to 0.65 V with a 10 nA range. An autosampler was programmed to mix a 100 μl sample aliquot with 50 μl of derivatization reagent containing 350 μg fluoraldehyde o-phthalaldehyde (OPA), 1% β-mercaptoethanol, in 0.4 M borate buffer, pH 10, 1 min before injection. The mobile phase contained 50 mM sodium acetate, 50 mM sodium phosphate adjusted to pH 6.8 with acetic acid, 2% tetrahydrofuran, 2% acetonitril, and 25% methanol. The flow rate was 1 ml/min and the total program time was 10 min. The column was rinsed with 80% methanol for 5 min and subsequently reequilibrated for 5 min between each sample run.
After deep anesthesia with equithesin (1 ml/kg), rats were fixed by an intracardiac perfusion with phosphate-buffered saline (PBS), pH 7.4, followed by 500 ml of fixative including 4% paraformaldehyde alone or 4% paraformaldehyde + 0.5% glutaraldehyde, in 0.1 M phosphate buffer, pH 7.4. After dissection, the brains were immersed for 2 to 3 days in the fixative without glutaraldehyde, then cut frontally into 40 μm thick sections with a vibratome. For the brains fixed with 4% paraformaldehyde alone, the sections were carefully rinsed in PBS and incubated for 48 h at 4°C with a mouse IgG monoclonal antibody against GFAP (Sigma; diluted 1:2,000) or rabbit IgG polyclonal antibody against glutamic acid decarboxylase (GAD; Chemicon; diluted 1:500). For the brains fixed with 4% paraformaldehyde + 0.5 glutaraldehyde, the sections were incubated with both the mouse IgG monoclonal antibody against GFAP and a rabbit IgG polyclonal antibody against GABA (diluted 1:5,000; kindly provided by Dr. Geffard, Bordeaux, France). After careful rinsing in PBS, sections were incubated for 2 h at 4°C with corresponding secondary antibodies conjugated to Cy3 (Jackson Laboratories, diluted 1:2,000, for single immunostainings) or to Cy3 and Alexa-fluor 488 (Molecular Probes). All primary and secondary antibodies were diluted in PBS containing 2% bovine serum albumin and 0.1% triton X100.
After decapitation of animals, the brains were dissected as indicated above and fixed by immersion in 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, for 2 to 3 days. The pieces of brain were then cut with a vibratome into 40 μm thick sections. After rinsing in PBS, these sections were either treated as described above for fluorescence immunostaining of GFAP or for macrophage marker ED1 (Chemicon, diluted 1:1,000), or mounted on slides and stained for 5–10 min in hematoxylin bath (Harris-type staining, Sigma).
After careful rinsing in PBS, immunostained sections were mounted with Mowiol (Calbiochem, La Jolla, CA) and observed under a Biorad MRC 1024 confocal laser scanning microscope. The background noise of each confocal image was reduced by averaging five image inputs. The organization of the immunostained structures was studied on reconstructed images obtained by collecting 5 to 15 consecutive confocal images 1 μm apart through the whole vibratome section thickness and by projecting on the same plane. Unaltered digitalized images were transferred to a PC type computer and used for the morphological analysis. For the different group of rats, the morphological organization of GFAP- or GAD-immunostained structures was analyzed within specific forebrain regions.
Ten μm thick reconstructed images obtained by using the × 20 objective were used to analyze the structural organization of GFAP-immunostained astrocytes. Within immunostained sections, 20 to 30 GFAP-immunostained (IS) astrocytes were analyzed within each of the forebrain regions considered, and three sections were used for three animals per treatment group. GFAP-IS astrocytes were classified into classes I to III, depending on the number of their primary and secondary processes [modified from the classification proposed by Mong et al. (2002); Fig. 5].
Five μm thick reconstructed images obtained by using the × 60 objective were used to evaluate the numerical density of GAD-IS axon terminals or perikarya within the different forebrain regions of untreated rats. For this, immunostained structures were counted within squared areas (150 μm side) centered on the forebrain regions considered, within three sections per animal and for three animals.
The specificity of the antibodies used has been previously assessed by the absorption test [see Damoiseaux et al. (1994) for the anti-ED1, Debus et al. (1983) for the anti-GFAP, Kaufman et al. (1986) for the anti-GAD, and Seguela et al. (1984) for the anti-GABA]. Additional controls consisted of omitting the primary antibodies and applying the secondary antibody alone.
All data reported in the present study are restricted to four regions of the adult rat forebrain, including the molecular layer of the hippocampus, the mediobasal hypothalamus, the dorsolateral striatum, and the medial layer of the parietal cortex.
Quantification of GFAP Protein and GABA Levels in Forebrain Regions of Untreated Rats
Tissue punch homogenates from different forebrain regions were analyzed for GFAP by Western blotting and for GABA by HPLC. Whatever the region considered, Western blots using the anti-GFAP antibody revealed only one immunoreactive band with an approximate molecular weight of 56 kDa. (Fig. 2B). The size and the optical density of the GFAP-immunoreactive band were found to vary considerably between the four forebrain regions analyzed. On the other hand, Western blots using an antibody against the housekeeping protein GAPDH revealed one immunoreactive band at 40 kDa, with a constant size and optical density from one forebrain region to the other (Fig. 2B). Quantification of the immunoreactive signals by computerized densitometry confirmed that constant amounts of GAPDH were detected within each of the different tissue extracts, and decreasing amounts of GFAP were detected within the hypothalamus, the hippocampus, the striatum, and the cortex (Fig. 3A).
HPLC analyses of GABA levels in the protein-free supernatant of the same tissue extracts showed a pattern similar to GFAP, with decreasing levels of GABA being detected in the hypothalamus, the hippocampus, the striatum, and the cortex (Fig. 3B). When fitted with a straight line of best-fit, a significant correlation (ρ = 0.92) was shown between the concentrations of GFAP and of GABA measured in the same tissue samples (Fig. 3C).
Morphological Organization of Astrocytes and GABAergic Neurons
In order to visualize the morphological organization of astrocytes and GABAergic neurons in the forebrain regions of untreated rats, sections were immunostained for either the astrocytic marker GFAP and/or for GABA or GAD, the synthesis enzyme of GABA. In all the immunostained sections examined, the organization of the different types of immunostained structures conformed to previous descriptions: that GFAP immunostaining was associated with the cell bodies and processes of astrocytes dispersed throughout white and gray matter regions (Fig. 4A–D), and that GABA and GAD immunostainings were very similar and were associated with neuronal cell bodies localized to gray matter regions and to numerous axon-like dot structures dispersed throughout both gray and white matter regions (Fig. 4E–H). In all the untreated rats examined, the immunostaining patterns for GFAP and for GABA or GAD were found to vary from one region to the other.
The examination of sections immunostained for GFAP showed that the density in GFAP-IS cells and their morphological organization were markedly different in the hypothalamus and hippocampus on the one hand and the striatum and the cortex on the other hand; the hypothalamus and hippocampus were found to contain the highest density of GFAP-IS cells, which generally exhibited a stellate organization with a large number of radiating processes (Fig. 4A and B), whereas the GFAP-IS cells located in the striatum and cortex appeared more dispersed and exhibited a more or less elongated shape with only a small number of processes (Fig. 4C and D). The analysis of the percent distribution of classes of GFAP-IS cells confirmed these differences, with a predominance of class III astrocytes in the hypothalamus and hippocampus and of classes I and II astrocytes in the striatum and cortex (Fig. 5D).
The examination of sections immunostained for GABA or GAD also showed marked differences between the different forebrain regions considered. Whereas cell bodies immunostained for GABA or GAD-IS appeared dispersed throughout all these forebrain regions, clear regional differences were observed concerning the organization of immunostained dot structures: in all the rats examined, the numerical density in these axon terminal-like labeled structures was found to be far higher in the hypothalamus and hippocampus than in the cortex and striatum (Figs. 1, 4, and 5). Double immunostaining of sections for GFAP and GABA further indicated that in the different forebrain regions considered, GABA immunostaining was never detected within GFAP-IS structures, although intensely GABA-IS terminal-like dots were frequently found to surround closely GFAP-IS cell bodies and processes (Fig. 1).
Taken together, these data indicate that in the forebrain of adult rats, the regions that contain the highest concentrations of GABA also contain the highest concentration of GFAP, corresponding to the occurrence of astrocytes exhibiting a high degree of stellation. To test the hypothesis that GABA signaling can affect the morphological organization of astrocytes and their expression of GFAP within forebrain regions of the adult rat, we chronically increased GABAergic signaling in vivo by using two different paradigms.
Intraperitoneal Administration of AOAA
AOAA is a potent inhibitor of GABA transaminase, which is an enzyme present in both neurons and astrocytes that converts GABA into succinic semialdehyde following its internalization from the extracellular space. In a preliminary study, we determined that the administration of AOAA to adult rats induced a rapid increase in GABA concentrations measured within forebrain tissue, which was maximum after 4 days (about 100% increase) and persisted until 15 days of treatment. We further determined that in rats treated with saline or AOAA for 1, 4, 9, and 15 days, only scarce degenerating cells exhibiting condensed chromatin could be detected throughout the different brain regions examined on hematoxylin-stained sections, and the brain parenchyma was devoid of ED1-immunostained cells, indicating an absence of macrophage infiltration and/or microglial cells activation (data not shown). It was thus assumed that the AOAA treatment had no toxic effects susceptible to induce the death of any neural cells.
In rats treated with AOAA for 9 days, HPLC quantification of GABA within tissue punches indicated that the GABA concentration was markedly increased within all the forebrain regions of the AOAA-treated rats as compared with the saline-treated controls (Fig. 6B). The Western blot analysis of the protein extracts of the same tissue punches indicated that AOAA treatment induced a significant increase of GFAP in all the regions analyzed (Fig. 6A). In both controls and those treated by AOAA, a significant correlation was evidenced between the concentrations in GFAP and GABA measured in the same tissue punches (Fig. 6C).
The examination of brain sections immunostained for GFAP first revealed that in control rats, the morphology of GFAP-IS cells was very similar in brain tissue fixed by immersion following decapitation of the animal (the case of all treated animals), and in brain tissue fixed by intracardiac perfusion of fixative (the case of untreated animals). In the forebrain regions of rats receiving IP injections of saline, the morphological organization of GFAP-IS astrocytes was similar to that previously observed in untreated rats fixed by perfusion. In AOAA-treated rats, by contrast, the structural organization of GFAP-IS astrocytes was markedly modified as compared to control rats; the number of astrocytic processes and of their branches was markedly increased in all the forebrain regions considered. Such modifications were particularly spectacular in the cortex and the striatum in which numerous astrocytes exhibited a high degree of stellation (Fig. 7). This was confirmed by the morphological analysis of GFAP-IS astrocytes indicating that AOAA treatment induced both a significant increase of class III astrocytes and a decrease in class I astrocytes (Fig. 8A).
These data indicate that increasing the intracerebral concentration of GABA by blocking its degradation affects both the expression of GFAP and the structural organization of astrocytes within the different forebrain regions considered. However, the systemic administration of AOAA affects peripheral organs that may affect CNS functions. In the last series of experiments, we decided to manipulate GABAergic signaling locally within the forebrain.
Intracerebroventricular Administration of Muscimol
Muscimol, a potent agonist of GABAAreceptors, was infused chronically for 4 days into the brain lateral ventricle by the means of a cannula connected to an osmotic Alzet micropump. A drawback of this approach is that the implantation of the injection cannula into the brain produces a mechanical lesion, which by itself induces an activation of the astrocytes surrounding the lesion. In control rats receiving an intraventricular injection of saline, the cortical areas closely surrounding the lesion were characterized by the presence of numerous pyknotic cells exhibiting condensed chromatin (not shown), the occurrence of numerous ED1-immunostained macrophage-like cells (Fig. 9A), and an increased numerical density in astrocytes exhibiting both an hypertrophy of their cell body and processes and an increased immunostaining intensity for GFAP (Fig. 9C). In all control and muscimol-treated rats examined, however, these histological signs were no more detected at more than 300 μm from the lesion (Fig. 9E and G). In this series of experiments, the analyses of GFAP content and of astrocyte morphology were thus restricted to forebrain areas located at more than 500 μm from the lesion (Fig. 10).
The Western blot analysis of punch extracts showed that, compared with controls, muscimol induced a marked increase in GFAP concentration in all the forebrain regions ipsilateral to the injection, and to a lesser extent on the contralateral side (Fig. 10). The examination of sections immunostained for GFAP confirmed that the intraventricular administration of muscimol induced spectacular modifications of the morphological organization of astrocytes located in forebrain regions ipsilateral to the injection, with a marked increase in the number of astrocytic processes and of their immunostaining intensity. The analysis of the morphological organization of GFAP-IS astrocytes in the cortex and striatum confirmed that muscimol induced both a decrease of class I and a marked increase of class III astrocytes within these regions (Fig. 8).
In the developing brain, immature astrocytes constitute the so-called radial glia, which appear as bipolar elongated cells. In the course of their maturation, these radial glial cells undergo dramatic modification of both their morphological organization and phenotype: their shape changes from bipolar to stellate (Schmechel and Rakic, 1979; Voigt, 1989; Culican et al., 1990), while their expression of GFAP progressively increases (Bovolenta et al., 1984; Pixley and De Vellis, 1984; Lendahl et al., 1990; Sancho-Tello et al., 1995; Eliasson et al., 1999). In a series of previous in vitro and in vivo studies, evidence has been provided that such a differentiation of immature astrocytes is at least partly mediated by the inhibitory neurotransmitter GABA (Matsutani and Yamamoto, 1997; Mong et al., 2002). The data reported here indicate that in the adult rat brain, GABAergic signaling may also influence the morphological organization and the GFAP expression of mature astrocytes.
In agreement with previous studies (O'Callaghan, 1991; Martin and O'Callaghan, 1995), our data show that adjacent forebrain regions exhibit marked variations in their GFAP contents. A possible explanation for this could be that the regions analyzed have different astrocytic densities. Since the concentration of GFAP within astrocytic cell bodies is often below the threshold for immunocytochemical detection, sections immunostained for GFAP cannot be used for the determination of the numerical density of astrocytes. In a series of previous studies, estimations of the regional densities in astrocytes have been made by counting the cell bodies immunostained for S100β, a calcium-binding protein that is detected within the astrocytic cell bodies, whatever their expression of GFAP (Legrand and Alonso, 1998; Savchenko et al., 2000). These data indicate that a good correspondence effectively exists between the GFAP content measured here in the different forebrain regions and their numerical density in astrocytes. However, the ratio between the highest (hypothalamus and hippocampus) and the lowest (cortex and striatum) numerical densities in astrocytes was less than 2, whereas the ratio between the levels of GFAP measured here in these regions was more than 10. This implies that in addition to differences in the number of astrocytes, marked differences exist between these regions in the amount of GFAP per astrocyte. The present observations indicate that, throughout the brain, immunostaining for GFAP is mainly associated with astrocytic processes and that, within the different forebrain regions considered here, the content in GFAP determined by quantitative Western blotting closely relates to the number and the degree of branching of the astrocytic processes. This provides strong support to the idea that the expression of GFAP plays a major role in the formation and maintenance of astrocytic processes (Weinstein et al., 1991; Rutka and Smith, 1993; Toda et al., 1994).
A prominent finding of the present study is certainly that the amounts of GFAP measured within the different forebrain regions of the adult rat brain are strongly correlated to the levels of GABA measured in the same regions. A possible explanation is that most of the GABA measured is contained within astrocytic cytosol. The present immunocytochemical data, however, clearly show that GFAP-IS astrocytic structures never exhibited immunostaining for GABA, indicating that these cells only contain low intracytoplasmic GABA levels. On the other hand, our immunocytochemical data clearly indicate that the forebrain regions that contained the highest GABA levels were those that contained the highest densities of GABA- and GAD-IS axon terminals (i.e., the hypothalamus and the hippocampus). This strongly suggests that GABA levels measured here in the different forebrain regions mainly correspond to the neurotransmitter contained within GABAergic terminals innervating these regions. Since previous in vitro and in vivo studies have provided evidence that GABA affects immature astrocytes by increasing both their GFAP expression and their stellation (Matsutani and Yamamoto, 1997; Mong et al., 2002), a possible interpretation of our data is that the differences in the expression of GFAP and in the morphological organization of astrocytes in different forebrain regions of the adult rat at least partly result from differential GABAergic signaling in these regions.
To evaluate this hypothesis, we have used two different experimental approaches aimed at chronically increasing the GABA signaling in vivo. We first treated rats with AOAA, a blocker of GABA degradation that is known to increase the intracerebral concentrations of GABA (Moguilevski et al., 1992; Carbone et al., 2002). Our data confirm that IP administration of AOAA for 1 to 15 days induced a significant increase in the GABA concentration in all the forebrain regions analyzed. Interestingly, we have also shown that after 9 days of AOAA treatment, this increase in GABA concentration was correlated with an increase in GFAP content within all the forebrain regions considered. A possibility is that such an increase in GFAP content corresponds to reactive astrogliosis induced by neuronal cell death consecutive to AOAA treatment. In the present study, the possible occurrence of cell death within the different brain regions considered has been investigated by looking for both the presence of macrophages or activated microglial cells that are known to concentrate within brain regions containing degenerated cells, and pyknotic degenerating cells exhibiting a condensed chromatin (Li et al., 1998). Our data, however, clearly show that the number of ED1-IS macrophages or pyknotic cells was not increased within the brain parenchyma after AOAA administration, whatever the duration of the treatment. Our morphological study further shows that AOAA treatment induced spectacular modifications of the structural organization of GFAP-IS astrocytes, mainly including increase in the number and the branching degree of astrocytic processes in all these forebrain regions. These data, however, suffer from the shortcoming that systemic administration of AOAA probably affects the functions of peripheral organs, which may indirectly affect astrocytes. In a second series of experiments, GABAergic signaling was thus modified intracerebrally by infusing muscimol directly within the lateral ventricle. Again, a drawback of this approach is that the lowering of the injection cannula through the cortex induces an activation of astrocytes, which may affect the detection of possible effects of the injected drugs. However, our observations on rats receiving intraventricular infusion of either saline or muscimol clearly show that significant cell death characterized by the infiltration of macrophages and the detection of pyknotic cells was limited to the cortical areas closely surrounding the cannula implantation site. Therefore, it is assumed that the marked increases of GFAP levels detected here in a number of forebrain regions at distance from the injection site of muscimol do not correspond to an astroglial response to cell damages. Importantly, these effects were associated with spectacular modifications of the morphological organization of the GFAP-IS astrocytes in the cortex and the striatum, which were very similar to the modifications observed in the AOAA-treated rats. Taken together, these data strongly suggest that GFAP expression and the morphology of adult astrocytes can be affected by GABAergic signaling.
The mechanisms by which GABA may influence astrocytes in situ are unknown. A first possibility is that GABA signaling indirectly affects astrocytes by affecting a population of neurons that in turn release signals to astrocytes. Nonetheless, the idea of a direct effect of GABA on astrocytes is strongly supported by a series of previous findings indicating that astrocytes express high levels of GABAA receptors (Fraser et al., 1994, 1995; Rosewater and Sontheimer, 1994) and that GABA or muscimol affects the morphological organization of astrocytes and their expression of GFAP when applied on pure astrocyte cultures (Matsutani and Yamamoto, 1997; Mong et al., 2002). It is thus reasonable to assume that in regions receiving a dense GABAergic innervation, part of the GABA released by the terminals diffuses to the GABAA receptors of surrounding astrocytes. Interestingly, a direct innervation of astrocytes by GABAergic axons can also be considered. Indeed, a series of previous studies have provided evidence for the existence of direct synaptoidic innervation of glial cells by various axon types (Aoki, 1992; Paspalas and Papadopoulos, 1996), including GABAergic axons (Buijs et al., 1987; Mudrick-Donnon et al., 1993).
In all the species studied so far (including human), aging is associated with an activation of astrocytes, characterized by an increased expression of GFAP and a progressive hypertrophy of the cell bodies and processes (Kohama et al., 1995; David et al., 1997; Amenta et al., 1998; Legrand and Alonso, 1998; Unger, 1998). Since such modifications of the morphology of astrocytes during aging dramatically increase the glial embedment of neuronal structures, it is possible that they affect neuronal functions by altering their synaptic input and/or the diffusion of various molecules and neurotransmitters within the extracellular space (Kullberg et al., 1998; Sykovà et al., 1998; Bacci et al., 1999). Within specific brain regions such as the hippocampus and the cortex, a clear correlation has been established between such an activation of astrocytes and aged-related neurological disorders such as memory deficit (Soffié et al., 1999). To date, the factors that are responsible for such modification of astrocytes during aging are not known. Interestingly, the synthesis of GABA has been reported to increase with increasing age in various brain regions (Reinikainen et al., 1988; Marczynski et al., 1994; Marczynski, 1998). Ongoing studies will try to determine whether such modifications of GABA signaling play a role in the age-related actvation of astrocytes.
Confocal microscopy has been realized using the facilities of Centre Régional Imagerie Cellulaire (CRIC), Montpellier, France. The authors are grateful to Anne Duvoy and Evelyne Galibert for their excellent technical assistance and Vicky Tobin for reading the article.