The brefeldin A-inhibited GDP/GTP exchange factor 2, a protein involved in vesicular trafficking, interacts with the β subunits of the GABAA receptors

Authors


Address correspondence and reprint requests to Dr Angel L. De Blas, 3107 Horsebarn Hill Road, U-4156, Storrs, Connecticut 06269–4156, USA. E-mail: deblas@oracle.pnb.uconn.edu

Abstract

We have found that the brefeldin A-inhibited GDP/GTP exchange factor 2 (BIG2) interacts with the β subunits of the γ-aminobutyric acid type-A receptor (GABAAR). BIG2 is a Sec7 domain-containing guanine nucleotide exchange factor known to be involved in vesicular and protein trafficking. The interaction between the 110 amino acid C-terminal fragment of BIG2 and the large intracellular loop of the GABAAR β subunits was revealed with a yeast two-hybrid assay. The native BIG2 and GABAARs interact in the brain since both coprecipitated from detergent extracts with either anti-GABAAR or anti-BIG2 antibodies. In transfected human embryonic kidney cell line 293 cells, BIG2 promotes the exit of GABAARs from endoplasmic reticulum. Double label immunofluorescence of cultured hippocampal neurons and electron microscopy immunocytochemistry of rat brain tissue show that BIG2 concentrates in the trans-Golgi network. BIG2 is also present in vesicle-like structures in the dendritic cytoplasm, sometimes colocalizing with GABAARs. BIG2 is present in both inhibitory GABAergic synapses that contain GABAARs and in asymmetric excitatory synapses. The results are consistent with the hypotheses that the interaction of BIG2 with the GABAAR β subunits plays a role in the exocytosis and trafficking of assembled GABAAR to the cell surface.

Abbreviations used
ARF

ADP-ribosylation factor

BIG

brefeldin A-inhibited GDP/GTP exchange factor

EM

electron microscopy

ER

endoplasmic reticulum

FNZ

flunitrazepam

GABAAR

γ-aminobutyric acid type-A receptor

GABARAP

GABAA receptor associated protein

GEF

guanine nucleotide exchange factor

GRIF-1

GABAA receptor interacting factor-1

GST

glutathione S-transferase

HA

hemagglutinin

HEK293

human embryonic kidney cell line 293

IL

large intracellular loop

mAb

monoclonal antibody

NMDA

N-methyl-d-aspartate

NSF

N-ethylmaleimide-sensitive factor

ORF

open reading frame

PKA

protein kinase A

SDS–PAGE

sodium dodecyl sulfate–polyacrylamide gel electrophoresis

TGN

trans-Golgi network

Y2H

yeast two-hybrid

Cytoplasmic proteins that interact with γ-aminobutyric acid type-A receptors (GABAARs) play a fundamental role in the trafficking and synaptic localization of these receptors. Thus, it has been proposed that gephyrin is involved in postsynaptic clustering of the GABAAR (Betz 1998). Gephyrin concentrates postsynaptically in inhibitory GABAergic synapses both in brain and in cultured neurons (Craig et al. 1996; Giustetto et al. 1998; Christie et al. 2002a,b). The GABAAR γ2 subunit and gephyrin are necessary for the clustering of both GABAAR and gephyrin at GABAergic synapses (Essrich et al. 1998; Kneussel et al. 1999). Nevertheless, a direct biochemical interaction between gephyrin and the GABAAR has not been demonstrated. Several groups have used the yeast two-hybrid assay (Y2H, Fields and Song 1989) to identify proteins that directly interact with GABAARs. Thus, GABAA receptor associated protein (GABARAP) was isolated using, as bait, the large intracellular loop (IL), located between TM3 and TM4 of the γ2S (γ2SIL) GABAAR subunit (Wang et al. 1999). GABARAP shows homology with light chain-3 of microtubule-associated proteins 1a and 1b and with GATE-16, a protein that facilitates transport of proteins through the Golgi apparatus via its interaction with the N-ethylmaleimide-sensitive factor (NSF, Sagiv et al. 2000). It has been proposed that GABARAP is involved in synaptic clustering of GABAARs (Chen et al. 2000) and/or GABAAR trafficking through the endoplasmic reticulum (ER) and the trans-Golgi network (TGN) (Kneussel et al. 2000; Kittler et al. 2001).

A GABAAR-interacting protein, Plic-1 (Bedford et al. 2001), was isolated by Y2H, using the IL of the GABAAR α1 subunit (α1IL) as bait. Plic-1 interacts with the ILs of all GABAAR α and β subunits. It has been proposed that Plic-1 competes with ubiquitin, preventing GABAAR poly ubiquitination and degradation. Another protein, GRIF-1 (GABAA receptor interacting factor-1), was also identified by Y2H using the β2IL as bait. GRIF-1 did not interact with any other GABAAR subunit ILs tested. A role for GRIF-1 in GABAAR intracellular trafficking has been proposed (Beck et al. 2002).

Brefeldin A is a fungal metabolite that inhibits protein secretion and induces the disintegration of the Golgi apparatus. BIG2 is one of several brefeldin A protein targets (Moss and Vaughan 1998). BIG2 (Morinaga et al. 1997; Togawa et al. 1999) is a Sec7 domain-containing guanine nucleotide exchange factor (GEF) that catalyzes GDP/GTP exchange on the small G-protein ADP-ribosylation factor 1 and 3 (ARF1/3). Activation of ARF1/3 by GDP/GTP exchange is required for membrane budding in the Golgi apparatus allowing proteins to progress through the TGN and the exocytotic pathway. It has been recently shown that a human mutation in BIG2 causes abnormal brain development, microcephaly and periventricular heterotopia (Sheen et al. 2004). In this communication, we show that, in the brain, BIG2 interacts with the β subunits of GABAARs. We originally revealed these interactions with Y2H. We have cloned and determined the full sequence of rat BIG2. Since the GABAAR β subunits are required for the translocation of assembled receptor from the ER to the plasma membrane (Connolly et al. 1996a,b), and since BIG2 is known to be involved in vesicular trafficking, we propose that the interaction of GABAAR with BIG2 plays a major role in the translocation of mature GABAAR assemblies, and likely other synaptic proteins, from the Golgi apparatus to endosomes and/or the plasma membrane. We also show for the first time the cellular and subcellular localization of BIG2 in the rat brain.

Materials and methods

All the animal protocols have been approved by the Institutional Animal Care and Use Committee and followed the National Institutes of Health guidelines.

Antibodies

All the anti-GABAAR and anti-BIG2 antibodies were raised in our laboratory. The mouse monoclonal (mAb) antiβ2/3 GABAAR subunit antibody (62–3G1) was raised to the affinity-purified bovine GABAAR (De Blas et al. 1988; Vitorica et al. 1988) and recognizes an N-terminal epitope that is common to the rat β2 and β3 subunits but is not present in the rat β1 subunit (Ewert et al. 1992). Both rabbit and guinea pig anti-rat α1 subunit antibodies were raised to amino acids 1–15 (QPSQDELKDNTTVFT) and the guinea pig anti-rat γ2 subunit antibody was raised to amino acids 1–15 (QKSDDDYEDYASNKY). The rabbit anti-β3IL antibody was raised to purified staphylococcal protein A (SPA)-β3IL fusion protein. All the rabbit and guinea pig antibodies used in light microscopy and electron microscopy (EM) immunocytochemistry were affinity-purified on immobilized peptide. The generation, affinity-purification, specificity and characterization of these antibodies have been described elsewhere (De Blas et al. 1988; Vitorica et al. 1988; Moreno et al. 1994; Miralles et al. 1999; Christie and De Blas 2003; Christie et al. 2002a,b).

For generation of BIG2-specific antibodies, the C-terminal synthetic peptide of the deduced amino acid sequence of rat BIG2 (PEEPSQVPAASTAW) was covalently coupled, via an N-terminal cysteine, to keyhole limpet hemocyanin. This peptide is not present in BIG1 (Togawa et al. 1999; Fig. 1b). A New Zealand rabbit was injected subcutaneously with a 1 : 1 emulsion of keyhole limpet hemocyanin-coupled peptide in Complete Freund's Adjuvant (for the first immunization) and with Incomplete Freund's Adjuvant (for all subsequent immunizations) once per month. The antibody titer in the serum was monitored by ELISA. Serum was collected after 4 months of immunizations and was affinity-purified on immobilized peptide. Anti-LexA mouse mAb was from BD Clontech (Palo Alto, CA, USA), anti-TGN38 mouse mAb was from BD Transduction Laboratories (San Diego, CA, USA), anti-hemagglutinin (HA) rabbit Ab was from Upstate Biotechnologies (Lake Placid, NY, USA) and anti-GABA guinea pig Ab was from Chemicon (Temecula, CA, USA). Fluorophore-labeled or colloidal gold-labeled secondary antibodies were from Jackson ImmunoResearch Laboratories (West Grove, PA, USA).

Figure 1.

Amino acid sequence alignment of rat and human BIG2 proteins. (a) Aligment of the deduced amino acid sequences of rat and human BIG2 (rBIG2 and hBIG2, respectively) from cDNA sequencing. The rat and human BIG2 proteins are 1791 and 1785 amino acids long, respectively. The Sec7 domain, from residues 648–834 of rat BIG2, is underlined. The 110 amino acid C-terminal fragment B91 originally isolated by Y2H, from residues 1679–1791, is in bold. Note the heptapeptide (residues 280–286, asterisks) that is present in rat BIG2 but absent in human BIG2. Also note that a residue present in human BIG2 at position 1781 is absent in the rat BIG2 (between position 1787 and 1788, asterisk). Dots indicate identical amino acids, uppercase letters indicate similar amino acids and lowercase letters indicate non-similar amino acids. Asterisks indicate that there is no corresponding amino acid in the homologous sequence. Hyphens indicate those gap positions. (b) Alignment of the C-terminus of rat BIG2 and rat BIG1. The amino acid sequence of the C-terminus of rat BIG1 was deduced from the published genomic sequence, after aligning the human BIG1 cDNA sequence (Togawa et al. 1999) to the rat genome database. The peptide used to generate a rabbit anti-rat BIG2 antibody, corresponding to residues 1778–1791, is underlined.

Yeast two-hybrid

All vectors and yeast strains for bait analysis and Y2H screening were from Dr Roger Brent (University of California, San Francisco, CA, USA) or Origene Technologies (Rockville, MD, USA). For bait construction, sense and antisense oligonucleotide primers were designed to amplify the IL of the various GABAAR subunits, such that each was directionally cloned into the polylinker of pEG202. For β3IL bait construction, a sense primer corresponding to the nucleotides 1052–1067 of the β3 subunit (Ymer et al. 1989) was designed with a BamHI restriction site at the 5′ end. An antisense primer, corresponding to nucleotides 1403–1420 of the β3 subunit (Ymer et al. 1989), was designed with a stop codon and the XhoI restriction site at the 5′ end. An identical procedure for bait construction was followed for the other GABAAR subunits except for α1, γ1IL and γ2IL where an EcoRI restriction site was added to the 5′ end of the IL instead of the BamHI site.

We confirmed that all LexA-bait fusion proteins were expressed in yeast by immunoblotting the cell lysate of yeast transformants with anti-LexA. We also confirmed that LexA-β3IL could not activate the LacZ reporter. For this purpose, Saccharomyces cerevisiae EGY48 was transformed with pSH18-34 and pEG202-β3IL and the LacZ reporter activity was tested by replica-plating the transformants on the appropriate X-GAL-containing media. For the positive control, the yeast was transformed with pSH18-34 and pSH17-4, and for a negative control, the yeast was transformed with pSH18-34 and pRHFM1. We also confirmed that the LexA-β3IL bait did not activate the genomic LEU2 reporter gene, since the pEG202-β3IL transformants did not grow in the absence of leucine.

For library screening, the yeast strain EGY48, previously transformed with pEG202-β3IL and pSH18-34, was transformed with pJG4-5 containing the oligo-dT primed rat brain cDNA library (Origene Technologies). An aliquot of the pooled transformants was then diluted 1 : 10 in liquid YNB medium containing galactose (to activate the GAL1 promoter) and allowed to incubate for 4 h at 30°C to induce cDNA library expression. Only clones with an activated leucine reporter grow on the medium lacking leucine. After allowing 4–6 days of growth, the fastest growing colonies were replica-plated onto solid YNB galactose growth medium containing X-GAL. Plasmids from yeast clones showing LacZ reporter activity (presumably expressing β3IL-interactors) were rescued by mechanical disruption and detergent lysis. The DNA was extracted with phenol/chloroform and was used to transform the trpEscherichia coli KC8 strain. Growth medium lacking tryptophan was used to select KC8 cells containing pJG4-5. Plasmid preparations from KC8 transformants were subjected to restriction analysis with EcoRI and XhoI, the enzymes used for directional cloning of the cDNA library into pJG4-5.

To map the binding site of the β3IL for the C-terminus of BIG2, various truncations of the β3IL cDNA (see Results section) were subcloned in pEG202, using a BamHI site at the 5′ end and a XhoI site at the 3′ end of each insert, and tested by Y2H as described above. To map the BIG2 binding site for the IL of the GABAAR β3 subunit, various truncations of the C-terminus of BIG2 (B91) were subcloned in pJG4-5, using a EcoRI site at the 5′ end and a XhoI site at the 3′ end of the insert, and tested by Y2H.

Full length cDNA cloning and sequencing of rat BIG2

The adaptor-ligated, Marathon-ready rat brain cDNA library (BD Clontech), containing full-length cDNA clones, was used as template. The 31-base sense primer corresponded to the 5′ end, starting at the putative ATG initiation codon, of the human BIG2 (Togawa et al. 1999). The 30-base antisense primer encoded the C-terminus of the rat BIG2 cDNA as we had previously determined by sequencing the rat B91 clone (the 110 amino acid C-terminal fragment of BIG2, originally isolated by Y2H, that interacted with the β3IL). The DNA sequence encoding the 19 C-terminal amino acids of BIG1 diverges from that of BIG2 (Fig. 1). Thus, these PCR primers specifically amplify the rat BIG2 and not BIG1. For PCR amplification, the Advantage 2 Polymerase Mix (Clontech) was used under the following thermocycling conditions: 94°C for 30 s followed by 30 cycles, each consisting of 94°C for 5 s and 70°C for 8 min. For cloning the full-length rat BIG2, the ∼5.6 kb fragment was bidirectionally ligated into pCR-TOPO-XL by T/A cloning with topoisomerase (Invitrogen, Life Technologies; Carlsbad, CA, USA). After transformation of E. coli TOP10 cells with the ligation product, three transformants were chosen for further analysis. The coding and non-coding strands of three of the resulting cDNA clones were independently sequenced using the BigDye Terminator DNA sequencing kit (Applied Biosystems, Foster City, CA, USA) and read using the ABI377 Prism DNA sequencer, model 377XL (Applied Biosystems). We have submitted the rat BIG2 cDNA sequence to GenBank (Accession number AY255526).

Preparation of crude synaptosomal/microsomal fraction

Forebrains or hippocampi from five 6–8-week-old male Sprague-Dawley rats were homogenized with a glass/Teflon homogenizer in 15 mL homogenization buffer (0.32 m sucrose, 10 mm Tris-HCl, pH 7.4, 1 mm EDTA, 1 mm phenylmethylsulfonyl fluoride, 10 µg/mL pepstatin, 10 µg/mL aprotinin A and 10 µg/mL leupeptin) at 4°C. The homogenate was centrifuged for 10 min at 1000 × g at 4°C and the supernatant was saved. The pellet was again suspended in 7.5 mL homogenization buffer and centrifuged as above. The two supernatants were pooled and centrifuged at 100 000 × g at 4°C, and the pellet was suspended in 7.5 mL lysis buffer (5 mm Tris-HCl, pH 7.4, containing the protease inhibitors 1 mm EDTA, 1 mm phenylmethylsulfonyl fluoride, 10 µg/mL pepstatin, 10 µg/mL aprotinin A and 10 µg/mL leupeptin), homogenized with a glass Dounce-homogenizer at 4°C and stored in aliquots at −70°C.

Coimmunoprecipitation of BIG2 and GABAAR from rat brain

For immunoblot analysis of proteins coprecipitated with anti-α1, 30 µL of protein-A Sepharose beads, suspended in 500 µL of 50 mm Tris-HCl, pH 7.5, were incubated with 2 µL of guinea pig anti-α1 antiserum overnight at 4°C with rotation. The crude hippocampal synaptosome/microsome fraction was incubated with an equal volume of 2 × RIPA buffer [20 mm Tris-HCl, pH 7.4, 274 mm NaCl, 2% deoxycholate, 2% Triton X-100 and 0.2% sodium dodecyl sulfate (SDS)] containing the aforementioned protease inhibitor cocktail at 4°C for 1 h. Detergent-insoluble material was pelleted by centrifugation at 100 000 × g for 1 h at 4°C. A volume of supernatant, containing 200 µg of protein, was added to the antibody-coated beads and incubated overnight at 4°C. Washed beads were incubated with SDS–polyacrylamide gel electrophoresis (SDS–PAGE) sample buffer (0.01 m Tris-HCl pH 6.8, 20% glycerol, 10%β-mercaptoethanol, 2.3% SDS, 0.005% bromophenol blue) for 20 min at room temperature followed by centrifugation. The supernatant was incubated in boiling water for 10 min and subjected to SDS–PAGE and immunoblotting with the rabbit anti-BIG2 antiserum. The immunoblotting procedure has been described elsewhere (De Blas and Cherwinski 1983).

For coimmunoprecipitation of the [3H]flunitrazepam ([3H]FNZ) binding activity of the GABAAR with anti-BIG2 antiserum, 1 mg/mL total protein of the crude forebrain (cerebral cortex and hippocampus) synaptosome/microsome fraction was solubilized with buffer A (30 mm CHAPS, 50 mm Tris-HCl, pH 7.4 and the aforementioned protease inhibitor cocktail) as described elsewhere (Mernoff et al. 1983) for 1 h at 4°C and centrifuged at 100 000 × g. Rabbit anti-BIG2 antiserum (50 µL) was added to 400 µL of the CHAPS-extract supernatant and incubated overnight at 4°C, followed by the addition of 30 µL of protein-A Sepharose and incubation for 1 h at 4°C. The beads were pelleted and the supernatant and pellet were recovered seperately to quantify GABAARs in each fraction by [3H]FNZ binding. For quantifying the immunoprecipitated GABAAR, 100 µL of buffer A were added to the pelleted beads together with 0.5 mL 50 mm Tris-HCl pH 7.4, and the slurry was incubated with 10 nm[3H]FNZ with or without 5 µm clonazepam (for displacement of specific [3H]FNZ binding, to determine non-specific binding) for 45 min at 4°C. To quantify the amount of non-immunoprecipitated receptor not adsorbed to the beads, 100 µL of the supernatant were incubated with 10 nm[3H]FNZ in 0.5 mL 50 mm Tris-HCl, pH 7.4, for 45 min at 4°C. Immunoprecipitated and non-immunoprecipitated GABAARs were incubated with 6.5% polyethylene glycol with 0.8 mg bovine γ-globulin as carrier at 4°C for 5 min, filtered through 24-mm Whatman GF/B glass fiber filters and washed with 5 mL 8% polyethylene glycol in 50 mm Tris-HCl, pH 7.4. Radioactivity was determined with a liquid scintillation counter.

Bacterial protein expression and filter overlay assay

The B91 cDNA was released from the Y2H prey vector pJG4-5 by digestion with EcoRI and XhoI and directionally cloned in the multiple cloning site of pET32a(+) (Novagen, Madison, WI, USA), such that the HIS tag was coupled to the N-terminus of the B91 cDNA insert. HIS-B91 expression was induced with 1 mm Isopropyl-β-D-thiogalactopyranoside (IPTG) in E. coli AD494(DE3) (Novagen) that had been transformed with the pET32a(+) containing the B91 cDNA insert. HIS-B91 was purified in batch from bacterial cell lysates with Ni2+-agarose beads, according to the manufacturer's instructions (Novagen). Bacterial expression and purification of glutathione S-transferase-β3IL (GST-β3IL) were done as reported elsewhere (Fernando et al. 1995). For the overlay assay, 1.25 µg HIS-tagged B91 was subjected to SDS–PAGE and transferred to nitrocellulose membrane (De Blas and Cherwinski 1983). Membrane strips containing immobilized HIS-tagged B91 were washed three times with Tris-buffered saline with Tween-20 (10 mm Tris-HCl pH 7.5, 200 mm NaCl, 0.2% Tween-20) for 5 min each and with Tris-buffered saline with Tween-20 containing 5% non-fat dried milk for 2 h. Strips were incubated with either 10 µg/mL GST-β3IL or 6.5 µg/mL GST (to maintain the same molar ratio) for 15 h, followed by three washes with 10 mm phosphate-buffered saline, pH 7.4, containing 0.05% Tween-20 for 5 min each. All the above steps were done at room temperature. The strips were incubated with phosphate-buffered saline with Tween-20 containing 5% non-fat dried milk for 30 min at room temperature followed by incubation with anti-GST antibody (Upstate Biotechnologies) diluted 1 : 2000 in phosphate-buffered saline with Tween-20 with agitation at 4°C overnight. Immunoreactive peptide bands were visualized as described above for immunoblots.

Low density hippocampal cultures

Hippocampal cultures were prepared by the method of Banker et al. (1998) as described elsewhere (Christie et al. 2002b). Briefly, dissociated neurons from embryonic day 18 Sprague-Dawley rat hippocampi were plated at low density (3000–8000 cells per 18-mm diameter circular coverslip) and maintained in glial cell conditioned medium for 24–31 days.

Transfections of HEK cells

Human embryonic kidney cell line 293 (HEK293) cells were cultured in high glucose Dulbecco's modified Eagle's medium (Gibco-BRL, Life Technologies) with 5% fetal bovine serum (from Gibco-BRL, Life Technologies, Gaithersburg, MD, USA), in a 5% CO2 atmosphere. HEK293 cells were cultured on poly-l-lysine-coated 18-mm cover slips and were transfected with pcDNA3.1(+)-HA-B91 alone or cotransfected with GABAAR subunits using Lipofectamine 2000 as described in the manufacturer's instructions (Invitrogen). Briefly, 1 µg of each plasmid DNA and 2.5 µL of lipofectamine 2000 were incubated in 100 µL Dulbecco's modified Eagle's medium at room temperature for 20 min followed by the addition of 400 µL of Dulbecco's modified Eagle's medium. Cells were washed twice with serum-free Dulbecco's modified Eagle's medium and then incubated with the transfection mixture for 5 h at 37°C in 5% CO2. The transfection mixture was removed and the cells were cultured in Dulbecco's modified Eagle's medium, 5% fetal bovine serum in 5% CO2 for 24–48 h at 37°C. The cDNAs for rat B91, BIG2 and the GABAAR subunits α1, β3 and γ2S were all in pcDNA3.1(+).

Immunofluorescence

Double or triple label immunofluorescence detection of various antigens with specific antibodies raised in various species was done as described elsewhere (Christie and De Blas 2002, 2003; Christie et al. 2002a,b). Briefly, coverslips containing primary hippocampal neurons or transfected HEK293 cells, were incubated in 4% paraformaldehyde and 4% sucrose in phosphate-buffered saline for 12 min at room temperature followed by cell permeabilization with 0.25% Triton X-100 in phosphate-buffered saline for 5 min and by treatment with 5% donkey serum in phosphate-buffered saline for 30 min at room temperature. The cultures were incubated with a mixture of the primary antibodies (defined in the legends for Figs 4 and 6) raised in various species, in 0.25% Triton X-100 phosphate-buffered saline for 2 h at room temperature, washed and incubated for 1 h at room temperature with a mixture of species-specific secondary antibodies all raised in donkey and conjugated to either Texas Red, FITC, or AMCA fluorophores (Jackson Immunochemicals, West Grove, PA, USA) in 0.25% Triton X-100 phosphate-buffered saline. Optimal primary antibody dilutions were determined by dilution series. For HEK293 cells, the cell nuclei were labelled with DAPI. The coverslips were washed with phosphate-buffered saline, and mounted using Prolong anti-fade mounting solution (Molecular Probes; Eugene, Oregon). Specificity of the immunolabelling was demonstrated by blocking the binding of the primary antibody with 20 µg/mL of the antigenic peptide. Moreover, no immunolabeling was obtained when the primary antibody was omitted. Images were collected using a 60 × pan-fluor objective on a Nikon Eclipse T300 microscope with a Sensys KAF 1401E CCD camera, driven by IPLab 3.0 (Scanalytics, Fairfax, VA, USA) acquisition software. Image files were then processed and merged for color colocalization using PhotoShop 4.01 (Adobe).

Figure 4.

Interaction of the β3 GABAAR subunits with the 110 amino acid C-terminus of BIG2 and the full-length BIG2 in HEK293 cells. Triple label immunofluorescence of transfected HEK293 cells. Nuclei were revealed by DAPI fluorescence (panels c, f both halves, l and o both halves). Panels (d)– (o) show examples of cells cotransfected with GABAAR subunit(s) and B91 or BIG2 compared to cells that were transfected only with GABAAR subunit(s). Sometimes the single and double-transfected cells were close enough such that they are displayed in the same field (g–i and j–l). More distant pairs of cells were brought near to each other in split panels (d–f and m–o, left and right halves). (a–c) HEK293 cells were transfected with only HA-B91 and protein expression was detected with the rabbit anti-rat BIG2 (a) and a monoclonal anti-HA antibody (b). (d–f) Cells were cotransfected with HA-B91 and the GABAAR β3 subunit cDNAs and protein expression was detected with the rabbit anti-BIG2 (d) and the mouse monoclonal anti-β2/3 antibody 62–3G1 (e), respectively. Split-panels in (d)– (f) show two HEK293 side by side: the cell in the left half was cotransfected with β3 GABAAR subunit and HA-B91, whereas the cell in the right half was only transfected with β3 GABAAR subunit. (g–i) Cells were cotransfected with HA-B91 and a combination of α1, β3 and γ2S GABAAR subunit cDNAs. The expressed proteins were detected with rabbit anti-BIG2 (g), guinea pig anti-α1 (h) and mouse anti-β2/3 monoclonal antibody (i). The cell on the bottom left of these panels was cotransfected with B91 and α1 and β3 GABAAR subunits whereas the cell in the upper right of these panels was not transfected with B91. (j–l) Cells were cotransfected with HA-B91 and γ2S GABAAR subunit cDNAs and the expressed proteins were detected with rabbit anti-BIG2 (j) and guinea pig anti-γ2 (k) antibodies. (m–o) Cells were cotransfected with full-length BIG2 and GABAAR β3 subunit cDNAs and the expressed proteins were detected with rabbit anti-BIG2 (m) and mouse anti-β2/3 monoclonal antibody (n). In (m)–(o), the split panels show a cell transfected with BIG2 and the β3 GABAAR subunit (left half) and another only transfected with β3 GABAAR subunit (right half). All cDNAs were from rat and the antibodies recognized the overexpressed rat B91 or rat GABAAR subunits. Anti-mouse, anti-rabbit or anti-guinea pig secondary antibodies were made in donkey and conjugated with FITC (b; d, right half; e, left half; h, k and n), Texas Red (a; d, left half; e, right half; g, j and m) or AMCA (I). The examples are representative from two (j–l), three (a–c, d–f, m–o) or four (g–i) experiments, with an average of 13 pairs of cells per experiment. Scale bar = 10 µm. See Results section for descriptions of the symbols and results.

Figure 6.

BIG2 is highly concentrated in the trans-Golgi network of cultured hippocampal neurons but also partially colocalizes with GABAARs in neuronal processes. Double label immunofluorescence with affinity-purified rabbit anti-BIG2 (a, c and e) and the mouse mAb TGN38 (b) or the mouse mAb 62–3G1 to the β2/3 subunits of the GABAAR (d and f). (a and b) BIG2 colocalizes with the trans-Golgi network marker TGN-38 in the soma (arrows) and proximal dendrites (arrowheads). (c–f) BIG2 (c and e) and GABAAR (d and f) immunofluorescence in dendrites. The rectangles in (c) and (d) are shown at higher magnification in (e)– (g). Some dendritic granules show both BIG2 (e) and GABAAR (f) immunofluorescence (arrows), as shown in the color overlay (g). Anti-BIG2 was detected with a Texas Red-conjugated donkey anti-rabbit secondary antibody and anti-TGN38 and anti-β2/3 were detected with a FITC-conjugated donkey anti-mouse secondary antibody. Scale bar = 20 µm (a–d) and 4.4 µm (e, f and g).

Light microscopy immunocytochemistry of rat brain sections

This procedure has been described elsewhere (De Blas 1984; De Blas et al. 1988; Moreno et al. 1994). Briefly, 60-day-old male Sprague-Dawley rats were perfused through the ascending aorta under anesthesia (80 mg/kg ketamine-HCl, 8 mg/kg xylazine, 2 mg/kg acepromazine maleate i.p.) with fixative consisting of 4% paraformaldehyde, 1.37% lysine and 0.21% sodium periodate in 0.1 m phosphate buffer, pH 7.4. Brains were frozen and sliced in parasagittal sections with a freezing microtome. Free-floating tissue sections were incubated for 24 h at 4°C with the affinity-purified rabbit anti-BIG2 antibody in 0.3% Triton X-100, 0.1 m phosphate buffer, pH 7.4. The washed tissue sections were incubated with a biotin-labeled anti-rabbit IgG and avidin–biotin–horseradish peroxidase complex (Vectastain Elite ABC Kit, Vector Laboratories, Burlingame, CA, USA). The reaction product was visualized by incubation with 3,3′-diaminobenzidine tetrahydrochloride in the presence of cobalt chloride and nickel ammonium sulfate. Sections were washed and mounted on gelatin-coated glass slides.

Preembedding electron microscopy immunocytochemistry

For Fig. 7(a) and Fig. 8(a and b) 34-day-old male Sprague-Dawley rats were anesthetized as described above and blood was flushed with 100 mL 0.12 m phosphate buffer, pH 7.2 at room temperature by intracardiac perfusion through the ascending aorta. Rats were then perfused with 200–300 mL of fixative (4% paraformaldehyde, 0.5% glutaraldehyde in 0.12 m phosphate buffer, pH 7.2). Parasagittal vibratome sections of 50 µm thickness were cut in phosphate-buffered saline at 4°C. Fixed brain sections were incubated with 3% normal goat serum in phosphate-buffered saline at room temperature for 1 h followed by incubation overnight, at 4°C, in affinity-purified rabbit anti-BIG2 primary antibody diluted 1 : 50 in phosphate-buffered saline. The antibody signal was detected using the ABC procedure (Vectastain Elite ABC kit, Vector Laboratories) as instructed by the manufacturer. Sections were washed three times with phosphate-buffered saline for 20 min each at room temperature followed by incubation with 3,3′-diaminobenzidine tetrahydrochloride/H2O2 solution for 5 min. Sections were washed for 20 min three to four times in phosphate-buffered saline at room temperature and then postfixed for 45 min to 1 h at room temperature with 1% osmium tetroxide in 0.12 m phosphate buffer, pH 7.2. Sections were washed four times for 15 min each with 0.12 m phosphate buffer, pH 7.2 and dehydrated with the following series of cold ethanol solutions: 50%, 70%, 85%, 95%, each three times for 5 min each, and in 100% ethanol three times for 15 min each. Dehydration was followed by infiltration in Polybed 812 resin with propylene oxide. Sections were embedded with Polybed 812 using aclar strips and polymerized at 60°C for 2 days.

Figure 7.

Subcellular distribution of BIG2 in neurons by EM immunocytochemistry. (a) Pre-embedding EM immunoperoxidase labeling with anti-BIG2 shows that immunolabeling accumulates (filled arrows) at the trans-Golgi network (filled arrowheads) and at the end of Golgi cisternae. (b–f) Postembedding EM immunogold labeling with anti-BIG2 (b–f) and with anti-β2/3 (b) GABAAR subunits. The double labeling experiment (b) shows colocalization of BIG2 and GABAAR (corresponding to 18 nm, filled arrow, and 10 nm, empty arrow, diameter gold particles, respectively) in some TGN vesicles and/or vesiculotubular structures (filled arrowheads, b). BIG2 immunogold labeling (arrows, c) is present at TGN cisternae (arrowhead, c) and in clusters associated with microtubules, both in the axon (d) and dendrites (e). Note the presence of myelin (arrowhead, d) wrapping the axon. BIG2 immunolabeling is also localized in the postsynaptic cytoplasm (f), associated with microtubules (arrowheads), near the postsynaptic density. The presynaptic terminal in (f) is labeled ‘T’. Gold particle size for (c), (d), (e) and (f) are 18 nm diameter. Scale bar = 83 nm (b) and 125 nm (a, c–f). The brain samples were from locus ceruleus (a), cerebral cortex (b and c) and cerebellum (d, e and f).

Figure 8.

Synaptic distribution of BIG2 in neurons by EM immunocytochemistry. (a and b) Pre-embedding EM immunoperoxidase labeling with anti-BIG2 shows immunolabeling at the postsynaptic density (arrows) and dendritic cytoplasm of an asymmetric type I synapse on a dendritic spine (a) and of a symmetric type II synapse on a dendritic shaft (b). Note that the immunoreaction extends from the postsynaptic density to the postsynaptic cytoplasm. (c–h) Postembedding EM immunogold labeling with rabbit anti-BIG2 (c–h) and guinea pig anti-GABA (f) or with a mouse monoclonal anti-β2/3 GABAAR subunits (g and h). BIG2 immunolabeling (arrows) is localized in the postsynaptic cytoplasm (c) and in the presynaptic boutons that contact both dendritic spines (Sp in d) and dendritic shafts (Sh in e). Double labeling experiments show that BIG2 immunolabeling (arrows, 12 nm diameter gold particles in f and 18 nm in g and h) is also localized presynaptically at GABAergic synapses identified not only by the symmetric synapse morphology but also by the presence of GABA in the presynaptic terminal (f, arrowheads, 18 nm gold particles) or GABAARs in the postsynaptic membrane (β2/3, g and h, arrowheads, 10 nm diameter gold particles). The brain samples were from hippocampus (a), locus ceruleus (b), cerebral cortex (c) and cerebellum (d–h). Presynaptic terminals are labeled ‘T’. Gold particle size for (c), (d) and (e) are 12, 18 and 18 nm diameter, respectively. In (c), (g) and (h), scale bar = 83 nm. In all other panels, scale bar = 125 nm.

Postembedding electron microscopy immunogold

For Fig. 7(d–f) and Fig. 8(c–h), the tissue preparation, freeze substitution, and postembedding immunogold labeling were done as reported previously (Riquelme et al. 2002). Briefly, 35–70-day-old male Sprague-Dawley rats were anesthetized as described above, and perfused with 60 mL Ringer's solution pH 6.9 at room temperature for 1 min followed by 800 mL of 4% paraformaldehyde/0.5% glutaraldehyde in 0.1 m phosphate buffer, pH 7.4 fixative. Sections 300–500 µm thick were cut with a vibratome and cryoprotected with 2 m sucrose and plunge-frozen in liquid propane cooled by liquid nitrogen (− 190°C). Samples were immersed in 1.5% uranyl acetate in anhydrous methanol at −90°C for 30 h. Samples were infiltrated with Lowicryl HM20 resin (Polysciences, Warrington, PA, USA) and polymerized with UV light (−45 to 0°C) during 72 h in a Leica AFS freeze substitution instrument (Leica, Vienna, Austria).

For all postembedding immunogold experiments of Fig. 7 and Fig. 8, we collected 70–80 nm-thick sections from the embedded tissue blocks on 400-mesh gold-gilded nickel grids, coated previously with a Coat-Quick ‘G’ pen (Electron Microscopy Sciences, Fort Washington, PA, USA), and performed a double-sided immunoreaction procedure (Matsubara et al. 1996) as described elsewhere (Riquelme et al. 2002). After incubations with a single (or a mixture of two for double labeling) affinity-purified primary antibody followed by colloidal gold labeled, species-specific secondary antibodies (or a mixture of two labeled species-specific anti-IgG secondary antibodies raised in the same species), tissue sections were counterstained with 2% uranyl acetate for 1 min and with Sato's lead solution (Sato et al. 1967) for 1 min, both at room temperature. No immunolabeling was observed when the primary antibody was omitted. For Fig. 7(b and c) the Lowicryl-emmbedded tissue blocks were kindly provided by Drs Zoltan Nusser and Peter Somogyi, who have described the embedding method elsewhere (Nusser et al. 1998). Briefly, rats were anesthetized with 220 mg/kg of pentobarbitone sodium (Sagatal). Rats were perfused with 4% paraformaldehyde, 0.1% glutaraldehyde and 0.2% picric acid in 0.1 m phosphate buffer, pH 7.4. Vibratome sections of 500 µm thick were cryoprotected in 2 m sucrose in phosphate buffer followed by slam-freezing, freeze-substitution with methanol at −80°C, and embedding in Lowicryl HM20 at −50°C. These tissue blocks were sectioned in our laboratory and the ultrathin sections were subjected to the same immunogold procedure described above.

Results

B91, a 110 amino acid C-terminal fragment of BIG2, interacts with the GABAAR β3IL in the yeast two-hybrid assay

The β3IL was subcloned into the multiple cloning site of pEG202 such that it was expressed as a fusion protein with the DNA binding protein LexA at its N-terminus. The fusion protein was used, as bait in the Y2H, to screen an adult rat brain cDNA library, for interacting proteins. Clone B91 was isolated from the library based on the strong interaction with the β3IL as revealed by the rapid emergence of strong blue color, when assayed for LacZ reporter activity. After isolation, expansion and sequencing of clone B91, it was determined to have an open reading frame (ORF) of 330 bp, encoding a 110 amino acid peptide. This protein fragment from rat was 92% identical and 95% similar to the analogous C-terminal peptide of the human BIG2 (Fig. 1a). The complete rat BIG2 sequence is 1791 amino acids in length (Fig. 1a). In the Y2H, B91 showed high binding affinity for the IL of all the GABAAR β subunits (β1IL, β2IL and β3IL). However, it did not interact with α1IL, γ1IL, γ2SIL, or γ2LIL (Fig. 2a).

Figure 2.

Mapping of the β3IL and BIG2 interaction sites. (a) The binding specificity of B91 for the ILs of several GABAAR subunits was tested in the yeast two-hybrid assay. B91 interacted with the ILs of β1, β2, and β3 but not with those of α1, γ1, γ2S or γ2L. All baits were tested for their expression by immunoblotting yeast cell lysate (not shown). (b) Several truncations of the β3IL were tested for their interaction with B91, the C-terminal fragment of BIG2, in the Y2H. Filled regions represent the amino acid sequences that are conserved in the β1IL, β2IL and β3IL. The amino acid residues present in each of the β3IL truncations are shown on the left. The amino acid residue numbers correspond to that of the whole β3 subunit. B91 interacted with the full β3IL bait (amino acids 328–450) and all the truncations that contained amino acid residues 328–345. B91 did not interact with either the LexA DNA binding protein alone or the unrelated drosophila bicoid protein bait. (c) Several truncations of the C-terminus of BIG2 (B91) were tested for their interaction with the β3IL. The amino acid numbers, corresponding to that of the full-length BIG2, are shown on the left. Y2H interactions are indicated by a ‘+’, where a ‘–’ indicates no interaction as determined by LacZ reporter activity.

Mapping of the β3IL and BIG2 binding sites in the yeast two-hybrid assay

In order to determine the region of the GABAAR β3 subunit that interacts with BIG2, various truncations of the β3IL were constructed in the Y2H bait vector and tested for interaction with B91, the BIG2110 amino acid C-terminal fragment originally isolated by the Y2H assay. As shown in Fig. 2(b), truncations 328–399, 328–360 and 328–345 interacted with B91. Each of these β3IL truncations have a common region of 18 amino acids (328–345) that is conserved among the ILs of the GABAAR β1, β2 and β3 subunits (Fernando et al. 1995) but that is not present in any of the other subunit isoforms. Therefore, we conclude that the 18 amino acid region that is localized at the N-terminus of the IL of the β3 subunit (NYIFFGRGPQRQKKLAEK) is involved in the interaction of the β subunits with BIG2.

Y2H clone B91, the C-terminus of BIG2 that interacts with the β3IL in Y2H, is a relatively small fragment (6.14%) of the full-length rat BIG2 protein. We constructed truncations of this fragment to determine the binding site for the β3IL. The interaction between B91 and β3IL was lost when either the N- or the C-termini of B91 were deleted (Fig. 2c) suggesting that the β3IL binding site on BIG2 is not restricted to a contiguous amino acid sequence but rather both the N- and C-termini of B91 are involved in the interaction with β3IL.

B91 interacts in vitro with the GABAAR β3IL

HIS-B91 was subjected to SDS–PAGE, immobilized on nitrocellulose membrane and overlaid with either GST-β3IL fusion protein or GST, followed by incubation with an anti-GST antibody to detect the binding of the GST fusion protein to the B91 peptide (Fig. 3a). The GST-β3IL bound to the immobilized 31 kDa B91 (Fig. 3a, lane 1), whereas GST did not bind to the B91 peptide (Fig. 3a, lane 2). An immunoblot of the B91 peptide with anti-BIG2 antiserum (Fig. 3a, lane 3) identified the mobility of B91 in the gel, confirming that the protein to which the β3IL bound was B91. This experiment showed that B91 and β3IL proteins interact not only in the Y2H assay but also in vitro.

Figure 3.

The 110 amino acid C-terminal fragment of BIG2 and the β3IL interact in vitro and the full-length endogenous BIG2 interacts with assembled GABAARs in the brain. (a) B91 (the 110 amino acid C-terminal peptide of BIG2) interacts with β3IL in a filter overlay assay. Filter-bound HIS-B91 fusion protein was overlaid with either GST-β3IL fusion protein (lane 1) or GST alone (lane 2) and probed with an anti-GST antibody, revealing that GST-β3IL (lane 1) but not GST (lane 2) binds to the immobilized 31 kDa HIS-B91. The 31 kDa peptide band in lane 1 comigrates with HIS-B91 (lane 3), as revealed with an anti-BIG2 antibody. (b) An antibody to BIG2 recognizes a brain protein of ∼200 kDa in immunoblots of a synaptosome/microsome fraction from rat forebrain (cerebral cortex and hippocampus). The immunoreactivity (lane 1) is blocked by 25 µg/mL of the antigen peptide (lane 2). (c) A guinea pig anti-rat α1 GABAAR subunit antibody coprecipitates BIG2 from detergent (RIPA) extracts of a rat hippocampal synaptosomal/microsomal fraction. Immunoblots with rabbit anti-BIG2 antiserum shows the presence of the ∼200 kDa BIG2 peptide in the precipitate (lane 1). This protein was absent in the precipitates with preimmune serum (lane 2). A protein of the same mobility is revealed in immunoblots of forebrain synaptosome/microsome fraction with anti-BIG2 (lane 3). (d) The anti-BIG2 antibody coprecipitates [3H]FNZ binding activity of CHAPS-solubilized GABAARs from a rat forebrain synaptosomal/microsomal fraction. Much of the total [3H]FNZ binding in the precipitate obtained with the anti-BIG2 antibody (histogram bar 1) was displaced by clonazepam (histogram bar 2), the latter representing the non-specific [3H]FNZ binding. The difference between histogram bars 1 and 2 represents the specific [3H]FNZ binding to precipitated GABAARs. No immunoprecipitation of GABAARs was observed when the anti-BIG2 antibody was incubated with blocking antigen peptide (histogram bar 3) or when the precipitation was done with preimmune serum (histogram bar 4) since the small amounts of [3H]FNZ binding observed in both cases were not displaceable by clonazepam (not shown). Values are mean ± SEM. The precipitated [3H]FNZ binding in histogram bar 1 is significantly different from the ones in histogram bars 2, 3 and 4 (*p < 0.05; **p < 0.01, n = 5). However, the values in histogram bars 2, 3 or 4 are not significantly different from each other. All the nitrocellulose strips within each panel (a–c) were from the same blotting experiment. Proteins were transferred to a nitrocellulose membrane and subsequently cut into strips. Each strip was subjected to different incubation conditions as described.

Generation of anti-rat BIG2-specific antibodies

A specific rabbit antibody to the 14 amino acid C-terminal peptide of rat BIG2 was generated. This peptide was chosen because it is very different from the corresponding C-terminal peptide or from any other region of rat BIG1 (Fig. 1b). No similar peptide was found in the protein databases. BIG1 and BIG2 are the two homologous ∼200 kDa GEFs that activate ARF1/3 (Togawa et al. 1999). The rat BIG1 cDNA has not yet been cloned. Nevertheless, we determined the C-terminal peptide of rat BIG1 (Fig. 1b) by aligning the published cDNA sequence of the human BIG1 with the rat genome sequence database (National Center for Biotechnology Information). The analysis showed that indeed there is a rat BIG1 gene homologous to the human BIG1, located in the rat chromosome 5, region q11. The analysis also showed that the C-terminal peptides of rat and human BIG1 (amino acids 1839–1849 of human BIG1) are identical. This is not the case for the C-terminal peptides of human and rat BIG2 as shown in Fig. 1(a).

Immunoblots of crude synaptosomal/microsomal fraction from rat forebrain (cerebral cortex and hippocampus) showed that the anti-BIG2 antibody recognized a ∼200 kDa protein band (Fig. 3b, lane 1), corresponding to the expected Mr for rat and human BIG2. The binding of the antibody to the ∼200 kDa protein was displaceable by the BIG2 antigen peptide (Fig. 3b, lane 2).

Native BIG2 and the fully assembled GABAARs interact in the rat brain

This was demonstrated by showing that BIG2 coprecipitated with an anti-GABAAR antibody and that GABAARs coprecipitated with an anti-BIG2 antibody. Immunoprecipitation with hippocampal GABAAR with guinea pig anti-α1 subunit antibody coprecipitated the ∼200 kDa BIG2 as immunoblots with the rabbit anti-BIG2 antibody showed (Fig. 3c, lane 1). We precipitated with anti-α1 instead of anti-β subunit antibody to demonstrate that BIG2 interacts with the assembled GABAAR heteropentamer, and not just with the unassembled GABAAR β subunits. The ∼200 kDa polypeptide band comigrated with the peptide band recognized by anti-BIG2 in immunoblots of rat hippocampal crude synaptosomal/microsomal fraction (Fig. 3c, lane 3). No precipitation of BIG2 occurred when preimmune guinea pig serum was used instead of the anti-α1 antibody (Fig. 3c, lane 2) or when the anti-α1 antibody was incubated with the α1 antigenic peptide (not shown).

In a reciprocal immunoprecipitation, rabbit anti-BIG2 antibody coprecipitated the assembled GABAAR heteropentamer from detergent extracts of forebrain crude synaptosomal/microsomal fraction, as determined by [3H]FNZ binding to GABAARs (Fig. 3d). Aproximately 10% of the solubilized GABAARs, measured as clonazepam-displaceable [3H]FNZ binding activity, were coprecipitated with the anti-BIG2 antibody, as shown by the binding of [3H]FNZ in the immunoprecipitate in the absence or presence of clonazepam (histogram bars 1 and 2, respectively). The difference of these values is defined as the specific binding of [3H]FNZ to the immunoprecipitated GABAARs. Incubation of precipitating antibody with BIG2 antigen peptide prevented the precipitation of GABAARs, as determined by the absence of specific [3H]FNZ binding as shown in histogram bar 3. In this case, the only binding detected was non-specific, since it was not displaceable by clonazepam (not shown). Similarly, the preimmune serum did not precipitate GABAARs since the [3H]FNZ binding (histrogram bar 4) was non-specific, that is not displaceable by clonazepam (not shown). These experiments support the notion that a subpopulation of GABAAR interacts with BIG2 in the brain. The results also indicate that BIG2 interacts with the fully assembled GABAAR since anti-BIG2 antibody can precipitate [3H]FNZ binding activity, which occurs only in fully assembled GABAAR containing α, β, and γ subunits (Smith and Olsen 1995; Sieghart and Sperk 2002). This notion is also consistent with the cotransfection studies of B91 (or BIG2) with GABAAR in HEK293 cells as shown in Fig. 4(g–i). Moreover, and as shown above, BIG2 coprecipitated with the anti-α1 GABAAR antibody. However, we have shown that the β subunit(s), but not the α1 subunit, interact with BIG2, also indicating that BIG2 coprecipitates with the assembled GABAAR containing both the α1, β subunits (and the γ2 subunit, since the GABAAR has [3H]FNZ binding).

As indicated above, the pool of GABAARs that coprecipitates with BIG2 is relatively small (∼10%). We propose that this represents the pool of assembled GABAARs present in intracellular vesicles that is in transit to the endosomes or to the neuronal plasma membrane.

B91 associates with GABAAR in host HEK293 cells

We have investigated whether B91 (with an HA epitope tag at the N-terminus) and the GABAAR β3 subunit associate in mammalian host cells (Fig. 4) using immunofluorescence with the anti-BIG2 antibody and antibodies to various GABAAR subunits. The results show that B91 associates with β3 (Figs 4d–f) or with β3-containing GABAARs (Figs 4g–i). However, B91 does not associate with γ2S (Figs 4j–l) or α1 (not shown). The anti-rat BIG2 antibody only recognized the transfected rat protein, since only human cells transfected with the rat HA-B91 cDNA showed immunoreactivity when immunolabeled with anti-BIG2 (Fig. 4a) and anti-HA (Fig. 4b). After DNA transfection of HEK293 cells, B91 expressed alone mainly accumulated in the nucleus (Figs 4a and b, arrows). Some B91 was also detected in the cytoplasm (Figs 4a and b, arrowheads). When either the β3 subunit (Fig. 4e, right half) or the β3-containing recombinant GABAAR was expressed in the absence of B91 (Figs 4h and i, arrowheads), the host cells exhibited, in both cases, intense immunofluorescence in the cell surface as well as in the cytoplasm and the perinuclear region (ER). Note that the cells transfected with GABAAR subunits shown in Fig. 4(e) right half and Fig. 4(h) upper right corner were not transfected with B91, as shown by the lack of anti-BIG2 immunoreactivity (Fig. 4d right half and Fig. 4g upper right corner). DAPI fluorescence was used to reveal the nuclei (Figs 4c and f both halves, Figs 4l and o both halves). These results are consistent with previous studies indicating that β3 homopentamers and α1β3γ2 heteropentamers are transported to the cell surface (Connolly et al. 1996a). However, when B91 and the GABAAR β3 subunit were coexpressed in the same cell, both B91 and the β3 subunit changed their normal subcellular distribution to colocalize in the same subcellular compartment(s), suggesting that they interact. In this case, coexpressed B91 and β3 mainly colocalized in cytoplasmic aggregates in the perinuclear region (arrows, Figs 4d and e, left halves). Although we do not understand the reason for the localization of B91 in the HEK293 cell nucleus when expressed alone, we used it to our advantage to reveal the interaction of B91 with the GABAAR subunits. Figure 4(g–i) shows that B91 also associates with the fully assembled β3 subunit-containing GABAAR when HEK293 cells are cotransfected with the recombinant GABAAR (α1, β3 and γ2S) and B91. In this case B91 also accumulated in the perinuclear region rather than in the nucleus (Fig. 4g, arrow). Moreover, the GABAAR containing both α1 (Fig. 4h, arrow) and β3 (Fig. 4i, arrow) subunit also accumulated at the perinuclear region colocalizing with B91. In contrast, cells expressing at least α1 and β3 (and probably γ2) but not expressing B91 (Figs 4g–i, arrowheads) exhibited both cytosolic and surface labeling. It is also worth noting that in Fig. 4(d–i), B91 had a dominant negative effect, blocking the translocation of the β3 homopentamer (Figs 4d and e, left halves) or the recombinant heteropentamer GABAAR (Figs 4g–i) to the cell surface, leading to the accumulation of the GABAAR in the perinuclear ER (arrows). This effect likely results from B91 blocking the interaction of the GABAAR with the endogenous BIG2 of the host cell. We also noted that in HEK293 cells coexpressing B91 and the β3 subunit, the heavy accumulations of these two proteins near one side of the nucleus changed the morphology of the nucleus in a manner that is related to the shape of the aggregate (arrows, Figs 4d–f, left halves). Figure 4(j–l) shows that B91 and the γ2 subunit do not interact, since in cells coexpressing B91 and the γ2S subunit, B91 remained in the nucleus (Fig. 4j, arrow) whereas the γ2S subunit accumulated in the cytoplasm (Fig. 4k, arrowhead).

Cloning and sequencing the full length rat BIG2

We cloned the full-length rat BIG2 cDNA by PCR as described in the Methods and sequenced both strands. The BIG2 cDNA was 5784 bp in length, compared with 5861 bp of the published human BIG2 sequence (Togawa et al. 1999; accession number NM_006420). The ORF for rat BIG2 was 5373 bp in length, compared with 5355 bp of the human BIG2 ORF. The ORFs of the rat and human BIG2 cDNAs have 88% nucleotide identity. The 3′ untranslated region (UTR) of rat BIG2 consisted of 408 bp, including an 18 bp poly(A) tail. The amino acid sequence of the translated ORF of rat BIG2 was determined to have a predicted molecular weight of 201.9 kDa and was found to be 97% similar to the amino acid sequence of human BIG2 (Fig. 1a) and 78% similar to the amino acid sequence of the human BIG1. When the complete sequence of the rat BIG2 cDNA was compared with the published rat genome sequence, we found that the BIG2 cDNA sequence aligned with 38 exons at region q42 on rat chromosome 3. Although the 5′ 31-base long sense primer used for the full-length cloning was based on the human BIG2 cDNA sequence, the corresponding bases were compared to the rat genome where it was determined that there was a 1-base difference that was translationally silent. Thus, within the region of the rat BIG2 polypeptide corresponding to the 5′ sense primer, the amino acid sequences of the rat and human BIG2 are identical. We have submitted the rat full-length BIG2 cDNA sequence to GenBank (Accession number AY255526). It includes the rat cDNA sequence corresponding to the 5′ sense primer.

Overexpression of BIG2 facilitates the exit of GABAAR (β3 subunit) from the ER

Whereas coexpression of B91 (the C-terminal fragment of BIG2) with the GABAAR β3 subunit (or recombinant GABAAR) had a dominant-negative effect, blocking the translocation of β3-containing GABAARs to the cell surface (as described above, Figs 4d–i), coexpression of the full-length BIG2 with β3 had the opposite effect, promoting the loss of the β3-containing GABAARs from the ER (Figs 4m–o), consistent with the notion that BIG2 facilitates the exit of GABAARs from the ER. In cells expressing the β3 subunit but not rat BIG2 (Fig. 4n, right half; see also Fig. 4e, right half), β3 concentrates on the cell surface (filled arrowhead) as well as in the perinuclear ER (arrow). When β3-containing GABAARs were coexpressed with the full-length rat BIG2 (Figs 4m and n, left halves), however, the perinuclear β3 immunofluorescence disappeared (Fig. 4n, left half, arrow), as revealed by nuclear DAPI stain (Fig. 4o, arrow). These results are consistent with the notion that the full-length BIG2 facilitates the translocation of β3-containing GABAARs from the ER to the cell surface. The immunofluorescence pattern of the transfected full-length BIG2 (Fig. 4m, left half, empty arrowheads) is consistent with its presence at the Golgi apparatus and/or the ER. There is little colocalization of the β3 subunit and BIG2 immunofluorescence in the cells cotransfected with β3 and full-length BIG2 (Figs 4m and n, left halves) while there is both surface (Fig. 4n, left half, filled arrowhead) and intracellularly localized β3. These results suggest that the interaction between the GABAAR β subunits and BIG2 is transient, promoting the rapid exit of the GABAAR pentamer from the ER and/or Golgi apparatus.

Immunocytochemical localization of BIG2 in the brain

Light microscopy immunocytochemistry of rat brain sections (Fig. 5a) showed that BIG2 is ubiquitously expressed in all regions of the brain, with the strongest immunoreactivity occurring in the neuropil of the corpus striatum (see also Fig. 5b), and the olfactory tuberculum, particularly in the Calleja Islands and the olfactory bulb (see also Fig. 5g). The perikaryon and proximal dendrites of neurons in various brain regions such as the pyramidal neurons of layer V (Fig. 5d) and other neurons in various layers (Fig. 5c) of the cerebral cortex, the pyramidal cells of the hippocampus (Fig. 5e) and the Purkinje cells of the cerebellum (Fig. 5f) were also strongly labeled. There was also immunolabeling of the neuropil throughout the brain. In the olfactory bulb (Fig. 5g), the mitral cells and the external plexiform layer were strongly labeled. The glomeruli and the granule cell layer of the olfactory bulb were also significantly labeled. The locus ceruleus was also strongly labeled (not shown). No tissue immunolabelling was detected when the primary antibody was incubated with 25 µg/mL of antigenic peptide or when the primary antibody was omitted (not shown).

Figure 5.

Light microscopy immunocytochemistry of BIG2 in the rat brain. (a) Parasagittal section of the rat brain immunolabeled with affinity-purified anti-BIG2 antibody. Immunoreactivity is ubiquitously distributed throughout the brain, although the highest concentration is located in the corpus striatum (CS) and the olfactory tuberculum (OT), particularly in Calleja's islands (CI). (b) Corpus striatum (CS), ependymal layer (EL) and the dorsal transitional zone (DTr). Note the high and diffuse labeling in the CS. (c and d) Layers I–IV and layers IV–VI of the cerebral cortex, respectively. Immunoreactivity is highest in the perikaryon and proximal dendrites of the pyramidal neurons of layer V. (e) The CA2 region of the hippocampus including the stratum oriens (SO), stratum pyramidale (SP), stratum lucidum (SL) and the stratum radiatum (SR). The perikaryon and proximal dendrites of the pyramidal cells of the SP are strongly labeled. (f) Cerebellum, molecular layer (ML), Purkinje cell layer (PK) and granule cell layer (GR). Note the strong labeling of the perikaryon of the Purkinje cells (PK). (g) Olfactory bulb, glomerular layer (GL), external plexiform layer (EP), mitral cell layer (ML) and granule cell layer (GR). Immunoreactivity is strong in mitral cells and EP but is also positive in the glomeruli and GR. Other abbreviations are CB, cerebellum; CC, cerebral cortex; HP, hippocampus; OB, olfactory bulb; TH, thalamus. Scale bar = 3 mm (a) and 100 µm (b–f).

Colocalization of BIG2 and GABAAR in cultured hippocampal neurons

Double label immunofluorescence experiments showed that BIG2 (Fig. 6a) was mainly localized in the TGN in the soma (arrows) and the proximal part of the dendrites (arrowheads), as shown by the colocalization of BIG2 (Fig. 6a) with the TGN marker TGN38 (Fig. 6b). In addition to the TGN, anti-BIG2 also labeled smaller and less intensely fluorescent granular structures, with diameters ranging from 165.6 nm to 423.2 nm, in the dendrites (Figs 6c and e), the smallest of which are likely to correspond with trafficking vesicles and the larger perhaps to synaptic structures. This interpretation is consistent with the EM data described below, showing the presence of BIG2 in vesicular structures in the perikaryon, dendrites and axons as well as in synapses. It is known that BIG2 and ARF1/3 are involved in vesicular trafficking (Kahn and Gilman 1984; Tsuchiya et al. 1991; Welsh et al. 1994; Moss and Vaughan 1995, 1998). Double label immunofluorescence experiments of BIG2 and GABAAR (Figs 6c–g) showed that some (12.3 ± 0.6% mean ± SEM; n = 3805 clusters, two experiments, 10 neurons) of these BIG2-containing vesicles (Figs 6c and e) were also labeled with GABAARs (Figs 6d and f), as shown by arrows and overlay (Fig. 6g), as revealed by the anti-β2/3 mAb 62–3G1 (Figs 6d and f). Conversely, 11.5 ± 1% (n = 4367 clusters) of the GABAAR clusters colocalized with BIG2. These experiments suggest that β2/3 and BIG2 colocalize in some intracellular trafficking vesicles in primary hippocampal cultures. This hypothesis was confirmed in the rat brain with double-label immunogold experiments at the EM level as shown below. Colocalization of GABAAR and BIG2 clusters is expected to be low since BIG2 is localized in intracellular vesicles and the majority of GABAAR clusters in these cells are localized at the cell surface (Christie et al. 2002b) and only ∼10% of GABAARs coprecipitate with BIG2 (Fig. 3d).

Subcellular distribution of BIG2 in neurons by EM immunocytochemistry

Preembedding immunoperoxidase (Fig. 7a) and postembedding immunogold (Figs 7b and c) experiments showed that BIG2 (filled arrows, Figs 7a–c) is localized in the TGN cisternae (filled arrowheads, Figs 7a–c), frequently at the end of the cisternae (Figs 7a and c). Double label immunogold experiments (Fig. 7b) showed colocalization of BIG2 (filled arrow) and GABAARs (empty arrow), corresponding to 18 nm and 10 nm gold particle diameter, respectively, on cytoplasmic vesicles or vesiculotubular structures (arrowheads, Fig. 7b). These results are consistent with the notion that the interaction between BIG2 and the GABAAR is involved in GABAAR trafficking. Moreover, Fig. 7(d and e) show clusters of BIG2 immunogold particles associated with microtubules in the axon (Fig. 7d), which was identified by the presence of myelin (arrowhead, Fig. 7d) and in the thick dendrites (Fig. 7e). These clusters of gold particles likely result from the presence of BIG2 in trafficking vesicles (that are not preserved by the tissue treatment i.e. fixation, freeze substitution and embedding, (Rubio and Wenthold 1999)) being transported along microtubules. BIG2 is also localized in the postsynaptic cytoplasm of symmetric synapses (Fig. 7f) associated with postsynaptic microtubules (Fig. 7f, arrowheads). The postsynaptic localization of BIG2 in symmetric synapses was also revealed by preembedding immunoperoxidase (filled arrow, Fig. 8b). BIG2 is also postsynaptically localized in asymmetric synapses as shown by pre-embedding immunoperoxidase (Fig. 8a) and postembedding immunogold (Fig. 8c) as indicated by arrows. Immunoreactivity was often found in the postsynaptic cytoplasm normally not associated with the postsynaptic membrane or density. BIG2 is also present in presynaptic terminals (arrows, Figs 8d–h) of both asymmetric (Figs 8d and e) and symmetric GABAergic synapses as shown in double label immunogold experiments of BIG2 (arrows) with presynaptic GABA (Fig. 8f) or postsynaptic GABAAR (Figs 8g and h) as shown by arrowheads. These results suggest that BIG2 might be involved in transport of GABAAR and other synaptic proteins from the TGN to the synapse.

Discussion

GABAARs are heteropentamers composed of combinations of various subunit classes and isoforms, including α1–6, β1–3, γ1, γ2 with two splice variants (a short, γ2S, and a long, γ2L, form), γ3, δ, ε, π and θ (Barnard et al. 1998; Sieghart and Sperk 2002). The majority of GABAARs in the brain contain α and β subunits combined with γ or δ subunits (Barnard et al. 1998; Sieghart and Sperk 2002). GABAAR pentamers are assembled in the ER and transported to the plasma membrane. GABAAR β subunits are required for translocation of assembled receptors from the ER to the plasma membrane (Connolly et al. 1996a; Connor et al. 1998). The presence of a given β subunit isoform may confer information regarding the routing of GABAARs to specific membrane domains (Connolly et al. 1996b). In host HEK293 cells, in the absence of other subunits, β3 forms homopentameric GABAARs that translocate to the cell surface (Connolly et al. 1996b; Taylor et al. 1999). It is thought that other α, β or γ subunits cannot form homopentamers and are retained in the ER. GABAAR subunits that do not assemble into pentameric receptors are quickly degraded (Connolly et al. 1996a). In this communication, we report that the C-terminus of BIG2 specifically interacts with the ILs of the three β subunit isoforms but not with that of the α and γ subunits tested. The conserved 18 amino acid sequence localized at the N-terminus of the IL of β subunits is involved in the interaction of these subunits with BIG2. The results presented here and the known role of BIG2 in vesicular trafficking are consistent with the hypothesis that the interaction of BIG2 with the β subunits of GABAARs is involved in translocation of assembled receptors across the TGN to endosomes and/or the plasma membrane.

BIG2 is a ∼200 kDa protein belonging to a class of high molecular weight GEFs that are sensitive to the fungal toxin brefeldin A and that possess a ∼200 amino acid Sec7 domain (Moss and Vaughan 1998). All Sec7 domain-containing GEFs catalyze GDP–GTP exchange, through their Sec7 domain activity, on members of the ARF family of small GTPases. ARFs are required for transport of proteins from a donor membrane to an acceptor membrane (Kahn and Gilman 1984; Moss and Vaughan 1995, 1998). BIG2 catalyzes the GDP–GTP exchange on class I ARFs (ARF1–ARF3) (Tsuchiya et al. 1991; Welsh et al. 1994). This reaction is involved in recruiting the clathrin/AP-1-coat complex to the TGN (Traub et al. 1993; Moss and Vaughan 1998; Donaldson and Lippincott-Schwartz 2000; Mellman and Warren 2000). Previous immunofluorescence studies using non-neuronal cell lines (i.e. HepG2) have shown that BIG2 concentrates in the TGN, colocalizing with the Golgi markers 58 kDa protein and AP-1, but also assumes a punctate distribution in the cytosol of the same cells (Yamaji et al. 2000; Zhao et al. 2002). In the present communication, we have reported the cellular and subcellular distibution of BIG2 in primary cultured neurons and in various regions of the brain. To the best of our knowledge, this is the first time that the cellular and subcellular localization of BIG2, not only in brain but also in any intact tissue, has been described. Although BIG2 mRNA is expressed at similar levels in various tissues including the brain (Togawa et al. 1999), the phenotype of a human mutation in BIG2 is most evident in the brain, causing microcephaly and periventricular heterotopia (Sheen et al. 2004).

The molecular interaction between BIG2 and the ILs of GABAAR β subunits, revealed in the present study, might play a role in the transport of newly assembled GABAARs by clathrin/AP-1-coated vesicles from the TGN to endosomes and perhaps to the synaptic plasma membrane. This hypothesis is supported by immunofluorescence and EM immunocytochemistry data showing colocalization of GABAARs and BIG2 in vesicles present in the soma and dendritic cytoplasm (Fig. 7b) and the localization of BIG2 in the postsynaptic cytoplasm of symmetric type II synapses (Fig. 7f and Fig. 8b). Moreover, in HEK293 cells, the dominant-negative effect of B91 (the C-terminus of BIG2 that interacts with the IL of β subunits and that lacks the Sec7 domain present in BIG2) blocks the transport of β3 homopentamers or β3-containing GABAAR heteropentamers to the cell surface (Figs 4d–i) while BIG2 promotes the loss of GABAARs from the ER (Figs 4m–o) which is consistent with the notion that BIG2 facilitates the exit of GABAARs from the ER. It has been reported that HEK293 cells express very low levels of the GABAAR β3 subunit (Ueno et al. 1996; Davies et al. 2000). If endogenous β3 is expressed in our HEK293 cultures, they are not detectable by the immunofluorescence assay. Thus, the results in Fig. 4 only illustrate the interactions of the overexpressed exogenous proteins.

Another vesicular trafficking complex, the Sec6/8-containing exocyst complex, is involved in delivery of NMDA receptors to the cell surface (Sans et al. 2003; Hoogenraad and Sheng 2003). This system does not seem to be involved in the trafficking of GABAARs (Sans et al. 2003). Therefore, the translocation of GABAARs and NMDA receptors to the cell surface may be mediated by different exocytotic systems.

The EM immunocytochemistry data suggest that vesicles containing BIG2 are transported to symmetric and asymmetric synapses both pre- and postsynaptically, where microtubules are sometimes observed (Fig. 7f, arrowheads), suggesting that they are transported by microtubule-based motor systems in axons (Fig. 7d) or dendrites (Fig. 7e). Thus, although BIG2 localizes at the TGN, it is unlikely that it is directly involved in the TGN sorting of proteins to either the axon or dendrites (Craig and Banker 1994; Allan et al. 2002) since it can be found at both.

Postsynaptic GABAARs do not seem to be the only synaptic proteins transported by BIG2-containing vesicles. This notion is supported by the presence of BIG2 in excitatory asymmetric synapses and by the known role of BIG2 and ARF1/3 in general vesicular trafficking. The presence of BIG2 in presynaptic terminals and in association with axoplasmic microtubules suggests that BIG2-containing vesicles are also delivered from the TGN to the presynaptic terminal. BIG2 could be present in transport vesicles that fuse with presynaptic endosomes and also in synaptic endosomes themselves, being involved in the formation of vesicles that bud from the endosomes. However, BIG2 does not seem to be involved in the fusion of synaptic vesicles with the presynaptic plasma membrane during transmitter release or in the recycling of synaptic vesicles from the plasma membrane to endosomes. Thus, it has been shown that ARF1/AP3 is involved in the formation of synaptic vesicles from endosomes, but the recycling of synaptic vesicles from the plasma membrane is ARF1/3-independent (Faundez et al. 1998; Zakharenko et al. 1999; Ryan 2001).

To date, little is known about the molecular mechanisms involved in the trafficking of GABAARs in the context of exocytosis. Most of what is known about GABAAR trafficking is related to endocytosis of cell surface GABAARs via clathrin/AP-2-coated vesicles (Tehrani and Barnes 1993, Tehrani et al. 1997; reviewed in Barnes 2001). Moreover, the ILs of GABAAR β and γ subunits associate with the clathrin adaptor protein AP2 (Kittler et al. 2000). Some internalized GABAARs are targeted to endosomes (Connolly et al. 1999a,b) whereas some are ubiquinated and degraded. The ILs of α and β subunits of internalized GABAARs can interact with the 67 kDa, ubiquitin-like protein Plic-1, which has been proposed to prevent ubiquination and degradation, thus creating an internal GABAAR pool that can be recycled to the plasma membrane (Bedford et al. 2001).

Besides BIG2, another protein, GABARAP, seems to play an important role in the exocytosis of GABAARs from the Golgi apparatus to the cell surface (Wang et al. 1999; Kittler et al. 2001). GABARAP (i) interacts with the γ2IL GABAAR subunit, (ii) is enriched in the TGN and (iii) binds to the N-ethylmaleimide-sensitive factor (NSF), a protein involved in cargo-containing vesicle trafficking and membrane fusion.

BIG2-activated ARF1 can stimulate phospholipase D and phosphoinositide 4-kinase activity, causing translocation of phospholipase D to cell membranes, including the plasma membrane, and could lead to membrane remodeling (Brown et al. 1993; Whatmore et al. 1996; Godi et al. 1999). Therefore, alterations in local membrane lipid composition induced by BIG2 activation of ARF1 might be involved in the transfer of GABAARs to the target membrane. In addition, BIG2 contains an A-kinase associated protein (AKAP) domain in its N-terminus, and protein kinase A (PKA) stimulation increases the proportion of BIG2 that associates with membranes (Li et al. 2003). Thus, exocytosis of GABAARs from the TGN to the cell surface might be regulated by PKA-dependent phosphorylation. It has been recently shown that the IL of β2 and β3 can directly bind A-kinase anchoring protein 150 (Brandon et al. 2003). PKA and other protein kinases can phosphorylate β3IL at S408 and S409, β1 at S409 and β2 at S410 (Moss et al. 1992a,b; McDonald and Moss 1997, McDonald et al. 1998). Nevertheless, these serines are not located in the interaction domain of the IL (amino acids 328–345) that interacts with BIG2, as shown above. Therefore, the phosphorylation of these serines by various protein kinases might not directly regulate the affinity of the interaction of GABAARs with BIG2. Nevertheless, the phosphorylation of these serines might regulate the interaction of GABAARs with other proteins that also could be involved in the exocytosis of GABAARs.

Acknowledgements

We would like to thank Dr Roger Brent for his generous gift of the yeast two-hybrid plasmids pEG202, pSH18-34, pJK101 and the Saccharomyces cerevisiae strain EGY48. We also thank Dr Peter Seeburg for the GABAAR subunit cDNA clones and Drs Peter Somogyi and Zoltan Nusser for providing the Lowicryl-embedded brain tissue blocks used in some EM immunogold experiments (Figs 7b and c). This work was supported by the National Institute of Neurological Disorders and Stroke grants NS38752 and NS39287.

Ancillary