Neuronal, Glial, and Epithelial Localizationof g-Aminobutyric Acid Transporter 2, a

High-Affinity g-Aminobutyric Acid PlasmaMembrane Transporter, in the Cerebral

Cortex and Neighboring Structures

FIORENZO CONTI,1* LAURA VITELLARO ZUCCARELLO,2 PAOLO BARBARESI,1

ANDREA MINELLI,1 NICHOLAS C. BRECHA,3,4 AND MARCELLO MELONE1

1Institute of Human Physiology, University of Ancona, Ancona, I-60020 Italy2Department of General Physiology and Biochemistry, Section of Histology and Human

Anatomy, University of Milan, I-20133 Milan, Italy3Departments of Neurobiology and Medicine, UCLA School of Medicine,

Los Angeles, California 900374Veterans Administration Medical Center, Los Angeles, California 90037

ABSTRACTNeuronal and glial high-affinity Na1/Cl 2-dependent plasma membrane g-aminobutyric acid

(GABA) transporters (GATs) contribute to regulating neuronal function. We investigated in thecerebral cortex and neighboring regions of adult rats the distribution and cellular localization of theGABAtransporter GAT-2 by immunocytochemistry with affinity-purified polyclonal antibodies thatreact monospecifically with a protein of 82 kDa. Conventional and confocal laser-scanning lightmicroscopic studies revealed intense GAT-2 immunoreactivity (ir) in the leptomeninges, choroidplexus, and ependyma. Weak GAT-2 immunoreactivity also was observed in the cortical paren-chyma, where it was localized to puncta of different sizes scattered throughout the radial extensionof the neocortex and to few cell bodies. In sections double-labeled with GAT-2 and glial fibrillaryacidic protein (GFAP) antibodies, some GAT-2-positive profiles also were GFAP positive. Ultrastruc-tural studies showed GAT-2 immunoreactivity mostly in patches of varying sizes scattered in thecytoplasm of neuronal and nonneuronal elements: GAT-2-positive neuronal elements includedperikarya, dendrites, and axon terminals forming both symmetric and asymmetric synapses;nonneuronal elements expressing GAT-2 were cells forming the pia and arachnoid mater; astrocyticprocesses, including glia limitans and perivascular end feet; ependymal cells; and epithelial cells ofthe choroid plexuses. The widespread cellular expression of GAT-2 suggests that it may have severalfunctional roles in the overall regulation of GABA levels in the brain. J. Comp. Neurol. 409:482–494, 1999. r 1999 Wiley-Liss, Inc.

Indexing terms: inhibitory transmission; neocortex; synapses; astrocytes; leptomeninges; confocal

microscopy

g-Aminobutyric acid (GABA) transporters (GATs) medi-ate high-affinity Na1/Cl2-dependent GABA uptake intoaxon terminals and glial processes (Iversen and Neal,1968; Iversen and Snyder, 1968; Neal and Iversen, 1969;Iversen, 1971; Iversen and Kelly, 1975; Tanaka and Bow-ery, 1996), thereby influencing neuronal excitability (Isaac-son et al., 1993; Mager et al., 1993) and the spread ofGABA to neighboring synapses (Krogsgaard-Larsen et al.,1987; Hamann and Rossi, 1997). Their function is notlimited to terminating the action of GABA, because theyalso may release GABA into the extracellular space(Schwartz, 1982; Pin and Bockaert, 1989; Attwell et al.,1993; Levi and Raiteri, 1993). To date, four GAT cDNAs

have been isolated from the rodent and human nervoussystems: GAT-1, GAT-2, GAT-3, and betaine-GABA trans-porter (BGT-1) (Guastella et al., 1990; Borden et al., 1992,

Grant sponsor: MURST; Grant number: COFIN97; Grant sponsor:University of Ancona Research Funds; Grant sponsor: CNR; Grant sponsor:National Institutes of Health; Grant number: EY04067; Grant sponsor:NATO; Grant number: CRG 960162.

*Correspondence to: Fiorenzo Conti, Istituto di Fisiologia Umana, Univer-sita di Ancona, Via Tronto, 10/A Torrette di Ancona, I-60020 Ancona, Italy.E-mail: f.conti@popcsi.unian.it

Received 31 December 1997; Revised 21 December 1998; Accepted 4February 1999

THE JOURNAL OF COMPARATIVE NEUROLOGY 409:482–494 (1999)

r 1999 WILEY-LISS, INC.

1994; Clark et al., 1992; Yamauchi et al., 1992; Liu et al.,1993), all with different pharmacological properties andtissue distributions (Borden, 1996).

Recently, in situ hybridization histochemistry and immu-nocytochemistry have been used to localize GATs in thecentral nervous system (CNS; Brecha and Weigmann,1994; Ikegaki et al., 1994; Durkin et al., 1995; Honda et al.,1995; Minelli et al., 1995, 1996; Johnson et al., 1996; Ribaket al., 1996; Conti et al., 1998a). In previous studies, wefound that, in the neocortex, where the role of GABAergicsynaptic transmission is important in both physiologicaland pathological conditions, GAT-1 is localized predomi-nantly to axon terminals forming symmetric synapses andto astrocytes (Minelli et al., 1995; Conti et al., 1998a),whereas GAT-3 is localized exclusively to astrocytes (Minelliet al., 1996).

Little is known, however, about the expression of GAT-2,a predicted 602 amino acid protein found in the brain andin nonnervous tissue (Borden et al., 1992; Ikegaki et al.,1994; Johnson et al., 1996; Howd et al., 1997; Obata et al.,1997). Whereas in vitro studies had shown high levels ofGAT-2 mRNA in some astrocytes (0–2A/type 2; Borden etal., 1995), initial immunocytochemical and in situ hybrid-ization studies suggested that GAT-2 is not expressed inthe brain parenchyma but is localized only to leptomenin-ges and ependyma (Ikegaki et al., 1994; Brecha et al.,1995; Durkin et al., 1995). However, because radioactive insitu hybridization techniques (Durkin et al., 1995) providepoor cellular resolution, and immunocytochemical investi-gations were concerned with the general distribution ofthe three GATs (Ikegaki et al., 1994; Brecha et al., 1995),the precise cellular localization of GAT-2 remains to beelucidated. In the present study, we used light and electronmicroscopic immunocytochemical techniques with affinity-purified antibodies (Johnson et al., 1996) to identify thecells that express GAT-2 in the cerebral cortex and neigh-boring regions.

MATERIALS AND METHODS

Adult albino rats (Charles River, Calco, LC, Italy) weigh-ing 180–250 grams were used in the present studies. Careand handling of animals were approved by the AnimalResearch Committee of the University of Ancona, in accor-dance with National Institutes of Health guidelines.

Antibodies

Two affinity-purified rabbit polyclonal antibodies (363Eand 365E; Johnson et al., 1996) directed to the predictedC-terminus (Borden et al., 1992) of rat GAT-2 (GAT-2594–602) were used for these studies. Although the two antibod-ies produced similar light and electron microscopic stain-ing patterns, antibody 365E was not suitable for Westernblot analysis. Accordingly, immunoblotting studies wereperformed by using antibody 363E.

Antibody characterization byimmunoblotting and immunoblocking

Four rats anesthetized with 30% chloral hydrate wereperfused transcardially with 4 mM cold Tris-HCl, pH 7.4,containing 0.32 M sucrose, 1 mM ethylenediamine tetraace-tic acid (EDTA), 0.5 mM phenylmethylsulphonyl fluoride(PMSF), and 0.5 mM N-ethylmaleimide (NEM). The neocor-tex was homogenized with a glass-Teflon homogenizer in10 volumes of ice-cold buffer (0.32 M sucrose; 4 mM

Tris-HCl, pH 7.4; 1 mM EDTA; and 0.25 mM dithiotreitol;Ikegaki et al., 1994). The homogenate was centrifuged at31,000g for 15 minutes at 4°C, the supernatant wasrecentrifuged at 3105,000g for 1 hour at 4°C, and thecrude membrane pellet (Thomas and McNamee, 1990) wasresuspended in homogenization buffer containing proteaseinhibitors (1 mM EDTA, 0.5 mM PMSF, 0.5 mM NEM) andeither used immediately or stored at 280°C. Proteinconcentrations were measured with the Bio-Rad proteinassay kit (Bio-Rad Laboratories, Hercules, CA; Bradford,1976). Aliquots of crude membrane fraction were mixedwith equal volumes of 2 3 electrophoresis sample bufferwith 4 M urea (final concentration) and subjected to 10%sodium dodecylsulfate (SDS) polyacrylamide gel electropho-resis with a 3% stacking gel under reducing conditions.The separated proteins were transferred electrophoreti-cally to a 0.45-µm nitrocellulose filter (Towbin et al., 1979).The blots were probed sequentially with the GAT-2 antibod-ies (dilution, 1:500) and goat anti-rabbit immunoglobulinG (IgG) conjugated to horseradish peroxidase and thenreacted with the BM chemiluminescence Western blottingkit (Boehringer Mannheim, Mannheim, Germany) follow-ing the manufacturer’s instructions. Chemiluminescentbands were recorded on X-Omat AR film (Eastman Kodak,Rochester, NY). For controls, blots were treated by usingthe same immunolabeling procedure except that the pri-mary antibody was either omitted or preadsorbed with1025 M rat GAT-1588–599 peptide, 1025 M rat GAT-2594–602peptide, or 1025 M rat GAT-3607–627 peptide.

Tissue preparation

For light microscopy, eight rats were deeply anesthe-tized with 30% chloral hydrate and perfused transcardi-ally with 0.1 M phosphate-buffered saline (PBS), pH 7.4,followed by 4% paraformaldehyde (PFA) in 0.1 M phos-phate buffer (PB), pH 7.4. For electron microscopy, fourrats were perfused with 4% PFA plus 0.5% glutaraldehydein PB. Brains were postfixed for 1–2 hours at 4°C in thesame fixative used for the perfusion, cut with a Vibratomein the coronal or parasagittal plane into 25–30-µm-thicksections that were collected serially in PBS, and stored at4°C until processing. Sections for electron microscopy werealways reacted within 24 hours from cutting. Unlessotherwise specified, data were collected from a region ofthe parietal cortex characterized by the presence of aconspicuous layer IV with intermingled dysgranular re-gions, densely packed layers II and III, and a relativelycell-free layer Va. This region corresponds to the primarysomatosensory cortex (SI; Zilles, 1985; Chapin and Lin,1990). At least five sections from each rat were processedfor each experimental procedure.

Procedure

For immunoperoxidase studies, free-floating sectionswere pretreated in 1% H2O2 for 30 minutes, rinsed in PBS,preincubated for 1 hour in normal sheep serum (NSS; 10%in PBS) with 0.3% Triton X-100, and then incubatedovernight at 4°C in GAT-2 primary antibody (dilution,1:1,000). Sections were rinsed in PBS, incubated for 15minutes in 10% NSS, then in biotinylated anti-rabbit IgG(BA1000; Vector Laboratories, Burlingame, CA) at a dilu-tion of 1:100 in PBS (1 hour at room temperature orovernight at 4°C), rinsed in PBS, incubated in avidin-biotin peroxidase complex (ABC; Hsu et al., 1981) for 30minutes, washed several times in PBS, and then incubated

GAT-2 IN THE CEREBRAL CORTEX 483

in 0.08% 3,38diaminobenzidine tetrahydrochloride (DAB)in Tris 0.05 M with 0.02% H2O2. Sections were thenwashed in PBS, mounted on gelatin-coated slides, airdried, dehydrated, and coverslipped. For immunofluores-cence, free-floating sections were incubated for 30 minutesin PBS containing 1% bovine serum albumin (BSA) and0.2% Triton X-100, then in GAT-2 antibodies diluted 1:250in PBS with 0.1% BSA (48 hours), and finally in fluoresceinisothiocyanate (FITC)-conjugated goat anti-rabbit affinity-purified secondary antibody (F0382; Sigma, St. Louis, MO)diluted 1:50 (1 hour). Other free-floating sections wereincubated in a mixture of GAT-2 and glial fibrillary acidicprotein (GFAP; mouse monoclonal, clone G-A-5; Bo-heringer Mannheim) antibodies. Primary antibodies werediluted 1:250 in PBS with 0.1% BSA. Sections weresubsequently incubated with a mixture of the appropriateaffinity-purified FITC- or tetramethylrhodamine isothio-cyanate (TRITC)-conjugated secondary antibodies (F0382and T7782, respectively; Sigma) that had been previouslyabsorbed with the IgG of the noncorresponding primaryantibody. Single- and double-labeled Vibratome sectionswere examined by using a TCS NT confocal laser-scanningmicroscope (Leica Lasertechnik GmbH, Heidelberg, Ger-many) equipped with a 75-mW krypton/argon mixed gaslaser. FITC and TRITC were excited with the 488 nm and568 nm lines, respectively; imaged separately; and mergedwith Leica Power Scan software. The fields of interest werescanned by using a 363 PL APO oil-immersion objectivewith a numerical aperture of 1.32. An acquisition time of450 lines per second, taking approximately 1 second toproduce a 512 3 512 pixel image, was selected. Theconfocal pinhole aperture varied from 70 µm to 95 µm. Toimprove signal-to-noise ratio, for each image, 16 frameswere averaged by Kalman filtering. Control experimentswith single-labeled sections confirmed that there was nocross-talk during simultaneous acquisition of FITC/TRITClabels.

For electron microscopic investigations, five Vibratomesections/rat were pretreated by using a mild ethanoltreatment (10%, 25%, 10%; 5 minutes each) to increase thepenetration of immunoreagents and processed by using animmunoperoxidase technique to detect GAT-2 immunore-activity (ir). GAT-2 antibodies were used at a dilution of1:1,000, and Triton X-100 was not used. After completion ofthe immunocytochemical procedure described above, sec-tions were washed in PB, postfixed for about 30 minutes in2.5% glutaraldehyde in PB, washed in PB, and postfixedfor 1 hour in 1% OsO4. The sections were dehydrated,cleared in propylene oxide, flat embedded in Epon-Spurrbetween acetate foils (Aclar; Ted Pella, Redding, CA), andpolymerized at 60°C for 36 hours. When polymerizationwas complete, embedded sections were examined under adissecting microscope. Small areas containing strips ofsomatic sensory cortex, leptomeninges, choroid plexus,and walls of the third ventricle were excised with razorblades and glued to cured resin blocks. Semithin (1 µm)sections were cut with a Reichert (Buffalo, NY, USA)ultramicrotome and collected on glass slides without coun-terstaining for light microscopic inspection. Ultrathin sec-tions were cut from the surface, counterstained with leadcitrate or left unstained, and examined with a Jeol T8electron microscope (Peabody, MA) or a Zeiss 902 electronmicroscope (Thornwood, NY). Identification of labeled andunlabeled profiles was based on established morphologiccriteria (Peters et al., 1991).

Immunocytochemical controls

Specificity of GAT-2 labeling was studied in the three(conventional, confocal, and electron microscopy) experi-mental series by using preimmune serum in place of theprimary antiserum and by preadsorbing the GAT-2 antise-rum overnight with 1025 M GAT-2594–602, GAT-1588–599, orGAT-3607–627 peptides.

RESULTS

Immunoblotting

The purified antibody against GAT-2 used for Westernblotting studies (363E) revealed a band of approximately82 kDa in the crude membrane fraction of cerebral cortex(Fig. 1A). Labeling of the blots was prevented by omissionof the primary antibody (not shown) or by preadsorptionwith GAT-2594–602 peptide (Fig. 1B). Preadsorption of theantibody with GAT-1588–599 and GAT-3607–627 peptides didnot prevent labeling of the blot (Fig. 1C,D).

Light microscopy: Distribution of GAT-2immunoreactivity

Diffuse and faint GAT-2 immunoreactivity was observedin brain parenchyma; it was most evident in the cerebel-lum followed by the brainstem, hippocampus, and cerebralcortex (not shown).

In the SI cortex, GAT-2 immunoreactivity was detectedin puncta and in few cell bodies and processes (Figs. 2, 3,5). Immunoreactive puncta were of different sizes andwere scattered throughout all cortical layers (Figs. 3A,4A), in the subcortical white matter (Fig. 5B), and underly-ing ependymal cells (Fig. 3D). They were also localized toglia limitans, which forms a continuous layer at the

Fig. 1. Specificity of g-aminobutyric acid (GABA) transporter 2(GAT-2) antibody assessed by immunoblotting (100 µg protein wereloaded for each lane). Immunoblotting with GAT-2 antibody (A), withGAT-2 antibody preadsorbed with 1025 M GAT-2594–602 (B), GAT-1588–599 (C), and GAT-3607–627 (D). The bars identify molecular weights of 93kDa, 66 kDa, and 43 kDa from top to bottom. Arrowhead indicates gelbottom.

484 F. CONTI ET AL.

cortical surface (Fig. 3A,B), and outlined blood vessels(Figs. 2A,B, 3A,B, 4A) and unlabeled cell bodies (Fig.3A,C). GAT-2-immunoreactive (ir) cells were present in thecortical parenchyma, mostly in supragranular layers andin layer VI (Figs. 2C–I, 5A), subcortical white matter (Figs.2L, 5B), and ependyma (Fig. 3D). GAT-2-ir cells wereheterogeneous in size and morphology (Fig. 2C–J); some ofthem appeared to be neurons (see, e.g., Figs. 2H,I, 5A),whereas others had the morphologic features of astrocytes(Fig. 2C,D,G). The same pattern of GAT-2 immunoreactiv-ity was observed in sections from different animals, with-out significant regional variations along the rostrocaudalor mediolateral extents of the neocortex.

Strong GAT-2 immunoreactivity was observed in theleptomeninges (Figs. 2A,B, 3A,B, 4A), where it was de-tected in the arachnoid, and along the thin arachnoidtrabeculae in the subarachnoid space (Figs. 2A, 3A,B, 5A).Several GAT-2-positive puncta were present throughoutthe cytoplasm of ependymal cells lining the walls of the

third ventricle (Fig. 3D), and intense GAT-2 immunoreac-tivity also was detected in the choroid plexus (Fig. 3E).

GAT-2 immunoreactivity was prevented when GAT-2antibodies had been preadsorbed with 1025 M GAT-2594–602(Fig. 4C), whereas no changes in immunoreactivity wereobserved in sections incubated with GAT-2 antibodiespreadsorbed with 1025 M rat GAT-1588–599 (Fig. 4D) orGAT-3607–627 (Fig. 4E) C-terminal peptides.

Light microscopy: GAT-2/GFAPcolocalization

To verify whether GAT-2 was expressed by astrocytes,we performed a double-labeling studies with antibodies toGFAP, an astrocytic marker (Bignami and Dahl, 1973). Insections double labeled for GAT-2 and GFAP, intense GFAPimmunoreactivity was found in astroglial cells and pro-cesses forming the glia limitans, surrounding blood ves-sels, and scattered in the cortical neuropil (Fig. 5A) and in

Fig. 2. A–J: Immunoperoxidase staining showing GAT-2 immuno-reactivity in the primary somatosensory cortex (SI) and neighboringregions of adult rats. Intense GAT-2 immunoreactivity is evident inleptomeninges (A) and along blood vessel walls (B). In the corticalparenchyma, GAT-2 immunoreactivity is present in puncta and in cellbodies of both glial cells (C,D,G) and neurons (E,F,H–J). Cells in C–H

are from supragranular layers, whereas those in I and J are from layerVI and subcortical white matter, respectively. Small arrows in Dindicate lightly stained glial cells. Scale bars 5 40 µm in B (alsoapplies to A); 100 µm in D (also applies to C); 10 µm in J (also applies toE–J).

GAT-2 IN THE CEREBRAL CORTEX 485

Fig. 3. Laser confocal images of single optical sections labeled forGAT-2. A,B: In sections of the SI cortex, intense GAT-2 labeling isevident in leptomeninges (L), arachnoid trabeculae (arrowheads), glialimitans (arrows), and the wall of a blood vessel (V). Small immunore-active puncta are scattered throughout the cortical neuropil. Curvedarrows in A indicate unlabeled cell bodies. C: GAT-2-immunoreactive

(ir) puncta of different sizes (arrows) in layer V. Asterisks indicateunlabeled cell bodies.D: Intense GAT-2 labeling is evident in ependy-mal cells (E) lining the third ventricle (V) and in puncta (arrows)scattered in the underlying neuropil. E: GAT-2 labeling in choroidplexus. Scale bars 5 50 µm in A; 5 µm in B,C; 12.5 µm in D; 25 µm in E.

486 F. CONTI ET AL.

the subcortical white matter (Fig. 5B). In all of theselocations, some of the GFAP-positive astrocytic processesalso contained GAT-2 immunoreactivity (Fig. 5A,B). It isnoteworthy that there was colocalization of GAT-2 andGFAP immunoreactivity in the glia limitans (Fig. 5A) andin some astrocytes, both in the cortical parenchyma (Fig.5A) and in the white matter (Fig. 5B). However, severalGAT-2-positive puncta and cell bodies were devoid of GFAPimmunoreactivity, and many astrocytic processes, identi-fied by their GFAP immunoreactivity, were not labeled byGAT-2 (Fig. 5A). In double-labeling immunofluorescenceexperiments, deletion of GAT-2 antiserum resulted in theabsence of green signal, whereas deletion of GFAP antise-rum resulted in absence of red signal.

Electron microscopy

Immunoreaction product indicating GAT-2 immunoreac-tivity was in the form of electron dense patches of variablesize, mostly scattered in the cytoplasm of several neuronaland nonneuronal profiles and, in fewer cases, apposed tothe inner plasma membrane of these elements (Figs. 6, 7).

Neuronal labeling. In all cortical layers, GAT-2 immu-noreactivity was present in perikarya (Fig. 6A), dendrites

of all calibers (Fig. 6B–D), axon terminals (Fig. 6F–I),myelinated axons (Fig. 6E), and occasionally spines (notshown). In all samples examined, the most common local-ization of GAT-2 immunoreactivity was dendritic.

In one case, we counted all GAT-2-ir neuronal elementsin one vertical strip of SI cortex. This analysis showedthat, out of 1,122 immunoreactive elements, 967 (86.18%;S.D., 6.34%) were dendrites, 91 (8.12%; S.D., 4.36%) wereaxon terminals, 51 (4.54%; S.D., 2.16%) were myelinatedaxons, and 13 (1.16%; S.D., 0.93%) were spines. In addi-tion, there were 287 astrocytic processes.

Labeled axon terminals were found in all cortical layers,but they were more frequent in deep layers than insuperficial layers. They were both of the asymmetric (Fig.6F) and symmetric (Fig. 6G) types, as inferred from thepresence or absence of a prominent postsynaptic specializa-tion, and preferentially contacted distal dendrites andspines. The localization of immunoreaction product inaxon terminals typically was distant from the synapticcleft, even in the cases in which reaction product wasapposed to the plasma membrane (Fig. 7F–I).

Glial labeling. In the cortical parenchyma, nonneuro-nal labeled profiles were astrocytic processes (Fig. 7). In

Fig. 4. Immunocytochemical controls. Adjacent sections processedwith GAT-2 antibody (A,B), with GAT-2 antibody preadsorbed with1025 M rat GAT-2594–602 (C), with GAT-1588–599 (D), and with GAT-3607–627 (E). GAT-2 immunoreactivity does not change in D and E but is

completely abolished in C, in which neither leptomeninges (arrows)nor blood vessel walls (asterisk) are labeled. Scale bars 5 100 µm in A;50 µm in E (also applies to B–D).

GAT-2 IN THE CEREBRAL CORTEX 487

Fig. 5. Analysis by confocal microscopy of the distribution of GAT-2(green signal) and GFAP (red signal) immunolabeling in the superfi-cial part of somatosensory cortex (A) and underlying white matter (B).The merged images are derived from 16 optical sections acquired atfocus intervals of 0.6 µm. Double labeling was checked in every singleoptical section. Yellow sites indicate coexistence of the two antigens. A:GAT-2 labeling (green) is evident in leptomeninges (L), arachnoid

trabeculae (T), wall of blood vessel (V), neuronal perikarya (C), and insmall puncta scattered in the neuropil; glia limitans (GL) and somecortical astrocytic processes show areas of colocalization of GAT-2 andGFAP (yellow). B: The subcortical white matter contains severalGAT-2-ir puncta (green) and double-labeled astrocytes (yellow). Scalebar 5 5 µm in A; 2.5 µm in B.

488 F. CONTI ET AL.

the large majority of cases, GAT-2 immunoreactivity wasin distal astrocytic processes (Fig. 7C,D), although, insome cases, proximal processes also were observed (Fig.7A). Intense GAT-2 labeling was present in processesforming glia limitans and in perivascular end feet profilesadjacent to the endothelial basal lamina (Fig. 7B), whereasendothelial cells always were unlabeled (Fig. 7B). Severalthin and irregularly shaped GAT-2-ir processes also werescattered throughout the cortical neuropil; in some cases,they were close to unlabeled axon terminals (Fig. 7C,D), inothers, they were found in the vicinity of labeled dendrites(Fig. 7C) but often in areas unrelated to synaptic termi-nals. Oligodendrocytes were not labeled.

In all locations that were studied at the electron micro-scopic level, GAT-2 immunostaining was not present insections processed with GAT-2 antibodies that had beenpreadsorbed with 1025 M GAT-2594–602, whereas immuno-staining was observed in sections incubated with GAT-2antibodies preadsorbed with 1025 M rat GAT-1588–599 andGAT-3607–627 C-terminal peptides.

DISCUSSION

Methodological considerations andcomparison with previous studies

Polyclonal antibodies developed to a synthetic peptidecorresponding to the predicted C-terminus of rat GAT-2were used to investigate the expression of GAT-2 in theneocortex and neighboring brain regions. These antibodieswere purified by using a peptide affinity column (Johnsonet al., 1996) and were characterized by using immunoblot-ting of crude membrane fractions of cerebral cortex (pre-sent study). These antibodies react monospecifically with aprotein of approximately 82 kDa, consistent with thepredicted molecular mass of the cloned GAT-2 (Borden etal., 1992). The specificity of these antibodies was demon-strated further by immunoblocking studies, which showedthat GAT-2 immunostaining is prevented when GAT-2antibodies are preadsorbed with the peptide used forantibody production, but not with other GAT C-terminalpeptides. Nevertheless, we cannot rule out the possibilitythat GAT-2 antibodies may have cross-reacted with someunidentified protein.

Of the three methods that can be used for the ultrastruc-tural immunocytochemical localization of antigens (i.e.,preembedding immunoperoxidase and pre- and postembed-ding immunogold), the preembedding immunoperoxidasemethod offers several advantages (for recent reviews, seeLujan et al., 1996; Ottersen and Landsend, 1997). First, itis universally considered the most sensitive procedure,thus allowing the visualization of low amounts of antigen;second, it allows the correlation between light and electronmicroscopic findings and therefore is the method of choiceto obtain information on the regional distribution of label-ing; finally, in our study, it was necessary so that we couldcompare the pattern of GAT-2 immunoreactivity with thepatterns described previously for GAT-1 and GAT-3 (Minelliet al., 1995, 1996). However, a major drawback of thepreembedding immunoperoxidase procedure is the diffu-sion of the electron-dense peroxidase reaction product.Although this feature hampers the precise subcellularlocalization of antigens, thus making any functional impli-cation quite speculative, it nevertheless does not raisedoubts about the cellular localization. Indeed, the localiza-tion of GAT-2 reported in this study can be regarded as

reflecting the actual presence of the antigen, because itwas confirmed by the immunofluorescence studies, and theperoxidase reaction product always was found intracellu-larly, in line with the location of the C-terminal portion ofGAT-2 to which antibodies were raised (Johnson et al.,1996), and it was absent in control sections.

Previous studies showed that, in the CNS, GAT-2 isexpressed exclusively in the leptomeninges and ependyma(Ikegaki et al., 1994; Brecha et al., 1995; Durkin et al.,1995). The present investigation extends previous resultsand shows that GAT-2 is expressed also by neurons andastrocytes. Given the specificity of the antibodies and thesensitivity of the immunocytochemical techniques used inthe present study, the poor cellular resolution of radioac-tive in situ hybridization techniques, and the fact thatprevious immunocytochemical investigations were gen-eral mapping studies and did not examine GAT-2 expres-sion at the subcellular and cellular levels (Ikegaki et al.,1994; Brecha et al., 1995), the low level of GAT-2 mRNA orprotein is likely to have passed unnoticed in earlier studies(Ikegaki et al., 1994; Brecha et al., 1995; Durkin et al.,1995).

Localization of GAT-2

Neuronal localization. The cardinal features of GAT-2expression in cortical neurons are 1) weak GAT-2 immuno-reactivity is present in dendrites, and 2) weak GAT-2immunoreactivity is localized to some axon terminals. Themost intriguing finding of the electron microscopic analy-sis was that, in all cases examined, GAT-2 immunoreactiv-ity was localized to dendrites, where neither GAT-1 norGAT-3 have been found (Minelli et al., 1995, 1996; Conti etal., 1998a). Even more puzzling was the observation thatGAT-2 immunoreactivity was in large patches in thecytoplasm of dendrites and not apposed to the innerplasma membrane, as, for example, in several immunore-active axon terminals (see Fig. 6F,H). Although we cannotrule out the possibility that GAT-2 might exert someunknown intracellular function, it is conceivable that themolecules that make up these patches of GAT-2 immunore-activity are proteins that are being transported to theplasma membrane, where eventually they will be inserted.The fact that we never observed GAT-2 immunoreactivityapposed to the inner plasma membrane of dendrites mayreflect the low numbers of molecules that are inserted atdifferent sites. If GAT-2 is inserted in the plasma mem-brane, then it might be either close or distant to regionscontacted by axon terminals. Clearly, the functional impli-cations of these two possible arrangements are quitedifferent: In the former, GAT-2 may function as a postsyn-aptic transporter, whereas, in the latter, it will not be inthe condition of modulating inhibitory transmission atsynaptic sites.

GAT-2 immunoreactivity also was observed in axonterminals forming both symmetric and asymmetric syn-apses. This dual localization appears typical of GAT-2,because it is the only known GAT expressed in axonterminals forming asymmetric synapses. Furthermore,although GAT-2 immunoreaction product was apposed tothe plasma membrane, it was never in areas associatedwith synaptic specializations, whereas, in GAT-1-ir axonterminals, reaction product always was close to the areasof plasma membrane associated with synaptic specializa-tion (Minelli et al., 1995; Conti et al., 1998a). Based onthese localizational features, it appears highly unlikely

GAT-2 IN THE CEREBRAL CORTEX 489

Figure 6

490 F. CONTI ET AL.

that the GAT-2 expressed by axon terminals may becapable of affecting the concentration of GABA in the cleftand, therefore, of interfering with inhibitory point-to-pointsynaptic transmission.

Overall, the present electron microscopic observationssuggest that GAT-2 might be involved more in regulatingGABA levels in the extracellular space than in the termina-tion of the synaptic action of GABA at the synaptic cleft.This postulated role for GAT-2, as well as for the other

GAT, is strengthened by the observation that GABA maydiffuse in the neuropil and act in a paracrine fashion(Dingledine and Korn, 1985; Thompson and Gahwiler,1992; Isaacson et al., 1993). Therefore, GAT-2 localized tononsynaptic neuronal sites may participate together withastrocytic GAT-2 (see below), GAT-1, and GAT-3 (Minelli etal., 1995, 1996) to the modulation of GABA’s diffuse action.This hypothesis also is consistent with the expression ofGAT-2 in axon terminals forming asymmetric synapses. Inthis regard, it is worth mentioning that the presence ofGAT-2 in a heterogeneous population of terminals is in linewith the results of previous studies, which showed GABAuptake in cortical glutamatergic, noradrenergic, dopamin-ergic, cholinergic, and peptidergic terminals (Bonanno andRaiteri, 1987a–c; Raiteri et al., 1991; Bonanno et al.,1993).

Nonneuronal localization. Immunolabeling forGAT-2 was found in several nonneuronal cell types, includ-ing cortical astrocytes and leptomeningeal, ependymal,and choroid plexus cells.

Astrocytic localization. The distinctive ultrastructuraland cytochemical features of astrocytes and astrocyticprocesses allow for their unambiguous identification (Pe-

Fig. 7. GAT-2 immunoreactivity in cortical astrocytes. A: Patchesof reaction product (long arrows) in a proximal astrocytic process closeto a labeled dendrite. Short arrows in the astrocytic process indicatebundles of gliofilaments. B: Labeled perivascular astrocytic process.C,D: Labeled distal astrocytic processes (arrows) in the vicinity of

unlabeled synapses. In C, the astrocytic process also is close to alabeled dendrite. Arrowheads indicate synaptic contacts. AsP, astro-cytic process; At, axon terminal; B, basal lamina; Cap, capillary; Den,dendrite; f, bundle of filaments; sa, spine apparatus; sp, spine. Scalebars 5 1 µm in A,D; 0.5 µm in B,C.

Fig. 6. GAT-2 immunoreactivity in cortical neurons. A: In peri-karya, reaction product (arrows) is associated with cisterns of endoplas-mic reticulum and Golgi apparatus. B–D: Patches of reaction product(arrows) in the cytoplasm of proximal (B) and distal (C,D) dendrites.Tailed arrows in B and D indicate a labeled and an unlabeledastrocytic process, respectively. In C and D, the labeled dendrites arecontacted by axon terminals forming symmetric synapses. E: A thin,myelinated axon containing a patch of reaction product (arrow).F–I: Patches of reaction product (arrows) are attached to the innerplasma membrane of an axon terminals forming either asymmetric (F)or symmetric (G) synaptic contacts. Arrowheads indicate asymmetric(B,F) and symmetric (C,D,G) synaptic contacts. AsP, astrocytic pro-cess; At, axon terminal; Ax, axon; Den, dendrite; Nuc, nucleus; sp,spine. Scale bars 5 1 µm.

GAT-2 IN THE CEREBRAL CORTEX 491

ters et al., 1991; Privat et al., 1995), and, by using thesecriteria, we have identified astrocytes and their processesin different preparations (for recent data from our labora-tory, see also Minelli et al., 1995, 1996; Conti et al., 1996,1998a,b). The specificity of the antibodies we used here,the conservative criteria adopted to identify astrocyticprocesses and perikarya, and the results of double-labelingstudies with the astrocytic marker GFAP strengthen theconclusion that some astrocytic processes in the neocortexof adult rats express GAT-2. With regard to their morphol-ogy and distribution, distal astrocytic processes labeled byGAT-2 are similar to those labeled with GAT-1 and GAT-3(Minelli et al., 1995, 1996). However, only GAT-2 immuno-reactivity is present in astrocytic cell bodies and theirproximal processes.

The existence of glial GABA uptake was demonstratedin the cerebral cortex in autoradiographic studies (Schonand Kelly, 1975). GABA transport in cortical astrocyticmembranes is highly sensitive to b-alanine (Mabjeesh etal., 1992), and sensitivity to b-alanine, but not to aminocy-clohexane, is a property that is associated with ‘‘glial’’transporters (Schon and Kelly, 1974; Iversen and Kelly,1975; Gavrilovic et al., 1984; Kanner and Bendahan, 1990;Mabjeesh et al., 1992). Furthermore, GABA uptake bycells transfected with GAT-2 or GAT-3 cDNA is stronglyinhibited by b-alanine, but not by aminocyclohexane (Bor-den et al., 1992; Clark et al., 1992; Clark and Amara,1994). The present observations and the results of therecent investigation of Borden et al. (1995), who showedthat GAT-2 mRNA is expressed in vitro by O-2A/type 2astrocytes, suggest that cortical glial GABA transport ismediated also by GAT-2.

Rat cortical astrocytes, therefore, express GAT-1 (Minelliet al., 1995; Conti et al., 1998a), GAT-2 (present study),and GAT-3 (Minelli et al., 1996), and BGT-1 has beenreported in type 1 astrocytic cultures derived from ratbrain (Borden et al., 1995), thus indicating that corticalastrocytes express all known GATs. These observationsraise several issues regarding the relative contribution ofeach of these transporters to overall GABA uptake by glialcells in the neocortex and of the functional significance ofmultiple GABA-uptake systems. In the absence of experi-mental studies, these issues can only be speculated upon.Based on the distribution of GATs (Minelli et al., 1995,1996; present study) as well as on their different distribu-tion and intensity of expression, it appears that 1) glialGABA uptake in the neocortex is mediated largely byGAT-3, and GAT-1 and GAT-2 play lesser roles; and 2) allGATs contribute to the uptake of extrasynaptic GABA.However, because the three GATs exhibit different ionicdependencies and inhibitor sensitivities (Guastella et al.,1990; Borden et al., 1992; Clark et al., 1992; Keynan et al.,1992), and because they are modulated differentially(Gomeza et al., 1991; Corey et al., 1994; Quick et al., 1997),it is conceivable that their relative contribution to glialGABA uptake is regulated dynamically and differentially,thus providing for a great adaptability in the regulation ofGABA extracellular levels. Finally, the role of BGT-1 in theneocortex remains to be established.

Localization to other nonneuronal cells. The detectionof GAT-2 immunoreactivity in leptomeninges, ependyma,and choroid plexus adds to the long list of transmitters,modulators, receptors, and transporters that are localized

in these structures (Nilsson et al., 1992; Del Bigio, 1995;Tang and Sim 1997) or to others specialized nonneuronalcells (Rudnick, 1977; Balkovetz et al., 1989; Wade et al.,1996; Schroeter et al., 1997). In these locations, GAT-2 maytake up circulating GABA from the cerebrospinal fluid(CSF) and be involved in osmotic control between CSF andbrain (Ikegaki et al., 1994; Durkin et al., 1995; Borden,1996; Johnson et al., 1996; Ramanathan et al., 1997). It isnoteworthy that GABAA receptor sites have been detectedin the epithelium of the choroid plexus (Amenta et al.,1989), and GABAergic supraependymal nerve fibers areubiquitous in the ventricular walls, where they maymodify local CSF composition; influence ependymal secre-tion, cell shape, or ciliary activity; act on distant targetsthrough the CSF; or be receptive to the composition of theCSF (Harandi et al., 1986).

CONCLUSIONS

This study shows that GAT-2 is distributed more exten-sively than was demonstrated previously, thus suggestinga role for this transporter in GABA uptake in the nervoussystem. There is strong GAT-2 expression in leptomenin-ges and ependyma and weak expression of this GAT incortical neurons and astrocytes and in a variety of special-ized epithelial cell types. The widespread cellular expres-sion of GAT-2 suggests that it may have several functionalroles in the overall regulation of GABA levels in the brain.

ACKNOWLEDGMENTS

We thank Maria (Charo) del Rio for insightful commentsand helpful discussions and Mrs. M. Marelli for preparingultrathin sections. This work was supported by funds fromMURST (COFIN97 to F.C.; 40% to L.V.Z.) and by VACareer Scientist Funds.

LITERATURE CITED

Amenta F, Cavalloti C, Collier WL, Ferrante F, Napoleone P. 1989.3H-muscimol binding sites within the rat choroid plexus: pharmacologi-cal characterization and autoradiographic localization. Pharmacol Res21:369–373.

Attwell D, Barbour B, Szatkowski M. 1993. Nonvesicular release ofneurotransmitter. Neuron 11:401–407.

Balkovetz DF, Tiruppathi C, Leibach FH, Mahesh VB, Ganapathy V. 1989.Evidence for an imipramine-sensitive serotonin transporter in humanplacental brush-border membranes. J Biol Chem 264:2195–2198.

Bignami A, Dahl D. 1973. Astrocyte-specific protein and neuroglial differen-tiation. An immunofluorescence study with antibodies to the glialfibrillary acidic protein. J Comp Neurol 153:27–38.

Bonanno G, Raiteri M. 1987a. A carrier for GABA uptake exists onnoradrenaline nerve endings in selective rat brain areas but not onserotonin terminals. J Neural Transm 69:59–70.

Bonanno G, Raiteri M. 1987b. Coexistence of carriers for dopamine andGABA uptake on a same nerve terminal in the rat brain. Br JPharmacol 91:237–243.

Bonanno G, Raiteri M. 1987c. Regional selectivity of a gamma-aminobu-tyric acid-induced 3H-acetylcholine release sensitive to inhibitors ofgamma-aminobutyric acid uptake. J Neurochem 48:1454–1458.

Bonanno G, Pittalunga A, Fedele E, Fontana G, Raiteri M. 1993. Glutamicacid and g-aminobutyric acid modulate each other’s through heterocar-riers sited on the axon terminals of rat brain. J Neurochem 61:222–230.

Borden LA. 1996. GABA transporter heterogeneity: pharmacology andcellular localization. Neurochem Int 29:335–356.

Borden LA, Smith KE, Hartig PR, Brancheck TA, Weinshank RL. 1992.Molecular heterogeneity of the g-aminobutyric acid (GABA) transportsystem. J Biol Chem 267:21098–21104.

492 F. CONTI ET AL.

Borden LA, Dhar TGM, Smith KE, Branchek TA, Glucowski C, WeinshankRL. 1994. Cloning of the human homologue of the GABA transporterGAT-3 and identification of a novel inhibitor with selectivity for thissite. Recept Chan 2:207–213.

Borden LA, Smith KE, Vaysse PJ-J, Gustafson EL, Weinshank RL,Branchek TA. 1995. Re-evaluation of GABA transport in neuronal andglial cultures: correlation of pharmacology and mRNA localization.Recept Chan 3:129–146.

Bradford MM. 1976. A rapid and sensitive method for the quantitation ofmicrogram quantities of protein utilizing the principle of protein-dyebinding. Anal Biochem 72:248–254.

Brecha NC, Weigmann C. 1994. Expression of GAT-1, a high-affinitygamma-aminobutyric acid plasma membrane transporter, in the ratretina. J Comp Neurol 345:602–611.

Brecha NC, Johnson J, Chen T, Conti F, Minelli A, De Biasi S, Ribak C.1995. GABA transporter expression in the rat nervous system [ab-stract]. Presynaptic mechanisms of neurotransmission. p 124.

Chapin JK, Lin C-S. 1990. The somatic sensory cortex of rat. In: Kolb B,Tees RG, editors. The cerebral cortex of the rat. Cambridge, MA: MITPress. p 341–380.

Clark JA, Amara SG. 1994. Stable expression of a neuronal g-aminobutyricacid transporter, GAT-3, in mammalian cells demonstrates uniquepharmacological properties and ion dependence. Mol Pharmacol 46:550–557.

Clark JA, Deutch AY, Gallipoli PZ, Amara SG. 1992. Functional expressionand CNS distribution of a b-alanine-sensitive neuronal GABA trans-porter. Neuron 9:337–348.

Conti F, De Biasi S, Minelli A, Melone M. 1996. Expression of NR1 andNR2A/B subunits of the NMDA receptor in cortical astrocytes. Glia17:254–258.

Conti F, Melone M, De Biasi S, Minelli A, Brecha NC, Ducati A. 1998a.Neuronal and glial localization of GAT-1, a high affinity g-aminobutyricacid plasma membrane transporter, in human cerebral cortex: with anote on its distribution in monkey cortex. J Comp Neurol 396:51–63.

Conti F, De Biasi S, Minelli A, Rothstein JD, Melone M. 1998b. EAAC1, ahigh-affinity glutamate transporter, is localized to astrocytes andGABAergic neurons besides pyramidal cells in the rat cerebral cortex.Cerebral Cortex 8:108–116.

Corey JL, Davidson N, Lester HA, Brecha NC, Quick MW. 1994. Proteinkinase C modulates the activity of a cloned g-aminobutyric acidtransporter expressed in Xenopus oocytes via regulated subcellularredistribution of the transporter. J Biol Chem 269:14759–14767.

Del Bigio MR. 1995. The ependyma: a protective barrier between brain andcerebrospinal fluid. Glia 14:1–13.

Dingledine R, Korn SJ. 1985. g-Aminobutyric acid uptake and the termina-tion of inhibitory synaptic potentials in rat hippocampal slices. JPhysiol (London) 366:387–409.

Durkin MM, Smith KE, Borden LA, Weinshank RL, Branchek TA, Gusta-fson EL. 1995. Localization of messenger RNAs encoding three GABAtransporters in rat brain: an in situ hybridization study. Mol Brain Res33:7–21.

Gavrilovic J, Raff M, Cohen G. 1984. GABA uptake by purified rat Schwanncells in culture. Brain Res 303:183–185.

Gomeza J, Casado M, Gimenez C, Aragon C. 1991. Inhibition of high-affinity g-aminobutyric acid uptake in primary astrocyte cultures byphorbol esters and phospholipase C. Biochem J 275:435–439.

Guastella J, Nelson N, Nelson H, Czyzyk L, Keynan S, Miedel MC,Davidson N, Lester HA, Kanner BI. 1990. Cloning and expression of arat brain GABA transporter. Science 249:1303–1306.

Hamann M, Rossi DJ. 1997. Different role of neuronal and glial GABAtransporters in cerebellar slices. J Physiol (London) 501.P:13P.

Harandi M, Didier M, Aguera M, Calas A, Belin MF. 1986. GABA andserotonin (5-HT) pattern in the supraependymal fibers of the ratepithalamus: combined autoradiographic and immunocytochemical stud-ies. Effects of 5-HT content on [3H] GABA accumulation. Brain Res370:241–249.

Honda S, Yamamoto M, Saito N. 1995. Immunocytochemical localization ofthree subtypes of GABA transporter in rat retina. Mol Brain Res33:319–325.

Howd AG, Rattray M, Butt AM. 1997. Expression of GABA transportermRNAs in the developing and adult optic nerve. Neurosci Lett 235:98–100.

Hsu SM, Raine L, Fanger H. 1981. Use of avidin-biotin-peroxidase complex(ABC) in immunoperoxidase techniques: a comparison between ABC

and unlabelled antibody (PAP) procedures. J Histochem Cytochem29:557–580.

Ikegaki N, Saito N, Hashima M, Tanaka C. 1994. Production of specificantibodies against GABA transporter subtypes (GAT-1, GAT-2, GAT-3)and their application to immunocytochemistry. Mol Brain Res 26:47–54.

Isaacson JS, Solis JM, Nicoll RA. 1993. Local and diffuse synaptic actions ofGABA in the hippocampus. Neuron 10:165–175.

Iversen LL. 1971. Role of transmitter uptake mechanisms in synapticneurotransmission. Br J Pharmacol 41:571–591.

Iversen LL, Kelly JS. 1975. Uptake and metabolism of g-aminobutyric acidby neurons and glial cells. Biochem Pharmacol 24:933–938.

Iversen LL, Neal MJ. 1968. The uptake of [3H]GABA by slices of rat cerebralcortex. J Neurochem 15:1141–1149.

Iversen LL, Snyder SH. 1968. Synaptosomes: different populations storingcatecholamines and gamma-aminobutyric acid in homogenate of ratbrain. Nature 220:796–798.

Johnson J, Chen TK, Rickmann DW, Evans C, Brecha NC. 1996. Multipleg-aminobutyric acid plasma membrane transporters (GAT-1, GAT-2,GAT-3) in the rat retina. J Comp Neurol 375:212–224.

Kanner BI, Bendahan A. 1990. Two pharmacologically distinct sodium- andchloride-coupled high-affinity g-aminobutyric acid transporters arepresent in plasma membrane vesicles and reconstituted preparationsfrom rat brain. Proc Natl Acad Sci USA 87:2550–2554.

Keynan S, Suh Y-J, Kanner BI, Rudnick G. 1992. Expression of a clonedg-aminobutyric acid transporter in mammalian cells. Biochemistry31:1974–1979.

Krogsgaard-Larsen P, Falch E, Larsson OM, Schousboe A. 1987. GABAuptake inhibitors: relevance to antiepileptic drug research. EpilepsyRes 1:77–93.

Levi G, Raiteri M. 1993. Carrier-mediated release of neurotransmitters.Trends Neurosci 16:415–418.

Liu Q-R, Lopez-Corcuera B, Mandiyan S, Nelson H, Nelson N. 1993.Molecular characterization of four pharmacologically distinct g-amino-butyric acid transporters in mouse brain. J Biol Chem 268:2106–2112.

Lujan R, Nusser Z, Roberts JDB, Shigemoto R, Somogyi P. 1996. Perisynap-tic location of metabotropic glutamate receptors mGluR1 and mGluR5on dendrites and dendritic spines in the rat hippocampus. Eur JNeurosci 8:1488–1500.

Mabjeesh NJ, Frese M, Rauen T, Jeserich G, Kanner BI. 1992. Neuronaland glial g-aminobutyric acid transporters are distinct proteins. FEBSLett 299:99–102.

Mager S, Naeve J, Quick M, Labarca C, Davidson N, Lester H. 1993. Steadystates, charge movement, and rates for a cloned GABA transporterexpressed in Xenopus oocytes. Neuron 10:177–188.

Minelli A, Brecha NC, Karschin C, De Biasi S, Conti F. 1995. GAT-1, ahigh-affinity GABA plasma membrane transporter, is localized toneurons and astroglia in the cerebral cortex. J Neurosci 15:7734–7746.

Minelli A, De Biasi S, Brecha NC, Vitellaro Zuccarello L, Conti F. 1996.GAT-3, a high-affinity GABA plasma membrane transporter, is localizedto astrocytic processes, and it is not confined to the vicinity ofGABAergic synapses in the cerebral cortex. J Neurosci 16:6255–6264.

Neal MJ, Iversen LL. 1969. Subcellular distribution of endogenous and[3H]g-aminobutyric acid in rat cerebral cortex. J Neurochem 16:1245–1252.

Nilsson C, Lindvall-Axelsson M, Owman C. 1992. Neuroendocrine regula-tory mechanisms in the choroid plexus-cerebrospinal fluid system.Brain Res Rev 17:109–138.

Obata H, Shimada K, Sakai N, Saito N. 1997. GABAergic neurotransmis-sion in rat taste buds. Immunocytochemical study for GABA and GABAtransporter subtypes. Mol Brain Res 49:29–36.

Ottersen OP, Landsen AS. 1997. Organization of glutamate receptors at thesynapse. Eur J Neurosci 9:2219–2224.

Peters A, Palay SL, de F. Webster H. 1991. The fine structure of the nervoussystem: neurons and their supporting cells. New York: Oxford Univer-sity Press.

Pin J-P, Bockaert J. 1989. Two distinct mechanisms, differentially affectedby excitatory amino acids, trigger GABA release from fetal mousestriatal neurons in primary cultures. J Neurosci 9:648–656.

Privat A, Gimenez-Ribotta M, Ridet J-L. 1995. Morphology of astrocytes.In: Kettenmann H, Ransom B, editors. Neuroglia. New York: OxfordUniversity Press. p 3–22.

GAT-2 IN THE CEREBRAL CORTEX 493

Quick MW, Corey JL, Davidson N, Lester HA. 1997. Second messengers,trafficking-related proteins, and amino acid residues that contribute tothe functional regulation of the rat brain GABA transporter GAT-1. JNeurosci 17:2967–2979.

Raiteri M, Bonanno G, Fedele E, Fontana G, Gemignani A. 1991. Gamma-aminobutyric acid (GABA) stimulates somatostatin release followingactivation of a GABA uptake carrier located on somatostatin nerveendings of rat cerebral cortex. J Pharmacol Exp Ther 256:88–93.

Ramanathan VK, Brett CM, Giacomini KM. 1997. Na1-dependent g-amino-butyric acid (GABA) transport in the choroid plexus of rabbit. BiochimBiophys Acta 1330:94–102.

Ribak CE, Tong W, Brecha NC. 1996. The GABA plasma membranetransporters, GAT-1 and GAT-3, display different distributions in therat hippocampus. J Comp Neurol 367:595–606.

Rudnick G. 1977. Active transport of 5-hydroxytryptamine by plasmamembrane vesicles isolated from human blood platelets. J Biol Chem252:2170–2174.

Schon F, Kelly JS. 1974. The characterisation of [3H]GABA uptake into thesatellite glial cells of rat sensory ganglia. Brain Res 66:289–300.

Schon F, Kelly JS. 1975. Selective uptake of [3H]b-alanine by glia: associa-tion with the glial uptake system for GABA. Brain Res 86:243–257.

Schroeter S, LeveyAI, Blakely RD. 1997. Polarized expression of the antidepres-sant-sensitive serotonin transporter in epinephrine-synthesizing chromaffin-cells of the rat adrenal gland. Mol Cell Neurosci 9:170–184.

Schwartz EA. 1982. Calcium-independent release of GABA from isolatedhorizontal cells of the toad retina. J Physiol 323:211–227.

Tanaka C, Bowery NG. 1996. GABA: receptors, transporters and metabo-lism. Basel: Birkhauser.

Tang FR, Sim MK. 1997. Metabotropic glutamate receptor subtype-1a(mGluR1a) ir in ependymal cells of the rat caudal medulla oblongataand spinal cord. Neurosci Lett 225:177–180.

Thomas CT, McNamee MG. 1990. Purification of membrane proteins.Methods Enzymol 182:499–520.

Thompson SM, Gahwiler BH. 1992. Effects of the GABA uptake inhibitortiagabine on inhibitory synaptic potentials in rat hippocampal slicecultures. J Neurophysiol 67:1698–1670.

Towbin H, Staehelin T, Gordon J. 1979. Electrophoretic transfer of proteinsfrom polyacrylamide gels to nitrocellulose sheets: procedure and someapplications. Proc Natl Acad Sci USA 76:4350–4354.

Wade PR, Chen J, Jaffe B, Kassem IS, Blakely RD, Gershon MD. 1996.Localization and function of a 5-HT transporter in crypt epithelia of thegastrointestinal tract. J Neurosci 16:2352–2364.

Yamauchi A, Uchida S, Kwon HM, Preston AS, Robey RB, Garcia-Perez A,Burg MB, Handler JS. 1992. Cloning of a Na1- and Cl2-dependentbetaine transporter that is regulated by hypertonicity. J Biol Chem267:649–652.

Zilles K. 1985. The cortex of rat. Berlin: Springer-Verlag.

494 F. CONTI ET AL.