A Novel Serine Kinase with Specificity for β3-Subunits Is Tightly Associated with GABAA Receptors*

Tuning of γ-aminobutyric acid type A (GABAA) receptor function via phosphorylation of the receptor potentially allows neurons to modulate their inhibitory input. Several kinases, both of the serine-threonine kinase and the tyrosine kinase families, have been proposed as candidates for such a modulatory role in vivo. However, no GABAAreceptor-phosphorylating kinase physically associated with the receptor has been identified so far on a molecular level. In this study, we demonstrate a GABAA receptor-associated protein serine kinase phosphorylating specifically β3-subunits of native GABAA receptors. The characteristics of this novel kinase clearly distinguish it from enzymatic activities that have been shown so far to phosphorylate the GABAA receptor. We putatively identify this protein kinase as the previously described GTAP34 (GABAA receptor-tubulin complex-associated protein of molecular mass 34 kDa). Using expressed recombinant fusion proteins, we identify serine 408 as a major target of the phosphorylation reaction, whereas serine 407 is not phosphorylated. This demonstrates the high specificity of the kinase. Phosphorylation of serine 408 is known to result in a decreased receptor function. The direct association of this kinase with the receptor indicates an important physiological role.

The ionotropic GABA A 1 receptor is abundantly expressed in the mammalian brain, and its major function is to mediate fast synaptic inhibition (for reviews, see Refs. [1][2][3]. Binding of GABA to the extracellular domain of the receptor leads to the opening of an intrinsic ion channel followed by chloride influx, counteracting depolarization of the neuronal resting potential. Also, excitatory actions of the receptor depending on an altered chloride equilibrium potential have been reported (4,5). A hallmark of the GABA A receptor is the many clinically important ways of modulation (3,6).
Biochemical purification of the GABA A receptor (7) and clon-ing of the first two subunits (8) were followed by the description of so far 14 different mammalian subunits named ␣1-6, ␤1-3, ␥1-3, ␦, and ⑀. From these subunits, diverse heterooligomeric receptor subtypes are generated neuronally (9), bearing most likely a pentameric structure (10,11) equivalent to that of the related nicotinic acetylcholine receptor (12). As also demonstrated for other ion channels (13), GABA A receptor function can be regulated by phosphorylation. This provides a mean for fine tuning inhibitory neuronal inputs. A variety of studies has given evidence for the phosphorylation of the GABA A receptor by serine/threonine kinases such as protein kinase C (14 -19, 23), cAMP-dependent protein kinase (17, 20 -25), cGMP-dependent protein kinase (26,27), Ca 2ϩ /calmodulindependent kinase II (27), and also by the tyrosine kinase Src (28,29). These reports point to the intracellular loops of GABA A receptor ␤and ␥-subunits as major phosphorylation targets. The functional consequences of phosphorylation are either inhibition or enhancement of GABA A receptormediated chloride currents. The results obtained for the effect of a given kinase are sometimes conflicting, probably reflecting the complex nature of GABA A receptor regulation by phosphorylation in vivo.
Kinase activities copurifying with native GABA A receptors have previously been reported (30,31). A kinase physically associated with native GABA A receptors can be expected to be of major functional relevance in vivo, but the molecular identity of such a protein has not been described so far. This study was conducted to answer the question whether one of the recently identified GABA A receptor-associated proteins named GTAPs (32) is a receptor-associated kinase. Our results indeed imply the identity of GTAP34 with a serine kinase that phosphorylates GABA A receptor ␤3-subunits.

MATERIALS AND METHODS
Immunoprecipitation of GABA A Receptors-GABA A receptors were immunoprecipitated from Triton X-100-or Zwittergent 3-14-solubilized calf brain membranes as described previously (32). For precipitation, the monoclonal antibody bd24 (33,34) covalently coupled to protein A-Sepharose (bd24-beads) was used. With 10 l of packed bd24-beads, GABA A receptor containing approximately 1 g of ␣1-subunit was precipitated. This value was determined by comparing the staining intensity of the ␣1-subunit, migrating in SDS-PAGE (35) as a sharp 50-kDa band and identified by Western blotting, with known amounts of marker proteins in silver-stained gels. Since around 55 kDa several GABA A receptor ␤-subunits and also receptor-associated tubulin comigrated, the amount of ␤-subunits present could not be determined.
Phosphorylation of Immunoprecipitated GABA A Receptor-For standard assays, 5 l of bd24-beads obtained after immunoprecipitation were washed five times with 1 ml of ice-cold buffer 1 (25 mM Tris, pH 7.2, 10 mM MgCl 2 , 0.1 mM Na-EGTA, 1% (w/w) ␤-octyl glucoside). 35 l of buffer 1 were added, and the reaction was started by adding radiolabeled ATP to concentrations ranging from 50 to 400 M. The specific activity was between 1000 and 10,000 cpm/pmol using [␥-32 P]ATP (3000 Ci/mmol; Hartmann Analytic, Braunschweig, Germany). Incubation was done under gentle agitation at 24°C in a water bath for 30 min unless indicated otherwise. The reaction was stopped by adding SDS-PAGE sample buffer and heating to 95°C for 10 min. After 8 or 10% SDS-PAGE, 1 silver staining (36), and drying of the gel, phosphorylated proteins were detected by autoradiography using a Kodak BioMax HE enhancer screen. To quantify 32 P incorporation, phosphorylated protein bands were excised from gels and homogenized, and the radioactivity was determined by scintillation counting.
Phosphorylation of Intracellular Loop Fusion Proteins and Histone-After immunoprecipitation of GABA A receptors, washing with buffer 1, and adding 35 l of buffer 1 to 5 l of washed bd24-beads as described above, 2 g of intracellular loop fusion proteins or an equimolar amount of histone H1 (fraction III-S; Sigma) was added. Since most of the fusion proteins were stored in 8 M urea stock solutions, their addition led to urea concentrations during the phosphorylation assay of 0.13-0.3 M. No effect of these concentrations of urea on the kinase activity was apparent (data not shown). Reactions were started by adding radiolabeled [␥-32 P]ATP to a concentration of 50 M, followed by incubation unter gentle agitation in a 24°C water bath for 60 min. For determining the stoichiometry of phosphorylation in the course of a time dependence (see Fig. 2), modified conditions were employed, i.e. 5 l of washed bd24-beads in a final volume of 20 l with 1 g of MBP-␤3-(345-408) fusion protein and 400 M ATP. Assays were terminated by pelleting the bd24-beads with the bound kinase activity by a brief centrifugation and removing the supernatants containing the intracellular loop fusion proteins. The supernatants were either analyzed by SDS-PAGE and autoradiography or by spotting aliquots onto Whatman P82 filters, extensive washing of the filters with 0.85% ortho-phosphoric acid, and scintillation counting.
Phosphorylation Assays after High Salt Extraction of GABA A Receptor Immunoprecipitates-Immunoprecipitates bound to bd24-beads were extracted with buffer 1 containing 0.6 M NaCl (high salt extraction) essentially as described previously (32). To maximally maintain kinase activity for phosphorylation assays, the primary incubation of the immunoprecipitate in high salt buffer was shortened to 1 min, and the extracts were diluted immediately after extraction 3-fold with buffer 1, resulting in a NaCl concentration of 0.2 M. Buffer 1 containing 0.2 M NaCl was also used for the phosphorylation assays of the other probes to be compared with the high salt extract. The assay volume was 120 l for each probe: the nonextracted bd24-beads, the extracted beads (both 5 l of beads) and the high salt extract (extracted from 5 l of beads). 5 g of MBP-␤3-(345-408) fusion protein or 0.78 g of histone were used as kinase substrate. The probes were incubated for 120 min at 24°C in the presence of 50 M radiolabeled ATP. Analysis of the assay supernatants was performed as described above.
Gel filtration of high salt-extracted proteins was done in buffer 1 containing 0.6 M NaCl with a fast protein liquid chromatography system on a Superose 12 HR10/30 column (Amersham Pharmacia Biotech). Undiluted extract obtained from 140 l of bd24-beads was separated at a flow rate of 0.1 ml/min. Starting at the void volume, 0.3 ml of eluate fractions were collected. For phosphorylation assays, 25 l of each fraction were diluted with 50 l of buffer 1 containing 4 g of the MBP-␤3-(345-408) fusion protein. To monitor autophosphorylation of separated proteins, identical probes lacking the fusion protein were prepared. Assays were performed for 120 min in the presence of 50 M ATP, followed by 8% SDS-PAGE and autoradiography.
Phosphorylation Assays of Immunoprecipitated ␤2and ␤3-Subunits-Calf brain membranes were solubilized in 10 mM HEPES, 5 mM EDTA, 1% SDS at 95°C for 10 min. The solution was diluted 10-fold with 10 mM HEPES, 5 mM EDTA, 50 mM NaCl, and 1% Triton X-100, resulting in a SDS concentration of 0.1%. After centrifugation at 165,000 ϫ g for 30 min, the supernatant was precleared with protein A-beads for 90 min at 4°C. 6 ml of the precleared solution containing 6 mg of protein were incubated with a 5 g/ml concentration of either the rabbit antibody ␤2-(351-405) or antibody ␤3-(1-13) that recognizes ␤2and ␤3-subunits (11) overnight at 4°C. The antibodies were recaptured with 15 l of packed protein A-beads and washed five times with buffer 1. 30 l of 1:1 slurry protein A-beads were mixed with 90 l of high salt-extracted kinase (prepared in Zwittergent 3-14, extracted and diluted as described above) and incubated for 120 min at 24°C in the presence of 50 M radiolabeled ATP. Phosphorylated ␤2and ␤3-subunits were analyzed by SDS-PAGE and autoradiography.
Phosphoamino Acid Analysis-One-and two-dimensional analysis of radiolabeled phosphoamino acids were performed essentially as described (44). In brief, GABA A receptor immunoprecipitate was incubated in buffer 1 with 50 M ATP for 60 min, either alone or in the presence of MBP-␤3-(345-408) fusion protein. After SDS-PAGE of the whole immunoprecipitate or only the probe supernatant, silver staining, and autoradiography, protein was eluted from gel pieces containing labeled GABA A receptor subunits or labeled MBP-␤3-(345-408) fusion protein. Following acidic hydrolysis, amino acids were separated by two-dimensional thin layer chromatography and compared with phosphoamino acid standards.
Two-dimensional Phosphopeptide Maps-Phosphorylation of immunoprecipitated GABA A receptors and MBP-␤3-(345-415) fusion proteins was done as described above. bd24-beads containing the bound phosphorylated receptor were washed with Lys-C digestion buffer (25 mM Tris HCl, pH 8.5, 1 mM EDTA). The supernatant of the phosphorylation assays containing the loop fusion proteins was subjected to a chloroform-methanol precipitation (45). Both the phosphorylated receptors and the loop constructs were digested in 200 l of Lys-C digestion buffer containing 0.1 g of Lys-C (Roche Molecular Biochemicals) for 19 h at 37°C. Peptides were lyophilized twice in H 2 O and separated on cellulose thin layer plates in formic acid, glacial acetic acid, and H 2 O at a ratio of 50:156:1794, pH 1.9, 1.8 kV, for 15 min. and by ascending chromatography in isobutyric acid, pyridine, glacial acetic acid, 1-butanol, and H 2 O at a ratio of 65:5:3:2:29 in the second dimension (46).

Phosphorylation of Isolated GABA A Receptor by an Endogenous Kinase
Activity-GABA A receptors were immunoprecipitated from Triton X-100-solubilized calf brain membranes with bd24-beads. The immunoprecipitate was assessed for the presence of an associated kinase activity by incubation with [␥-32 P]ATP and subsequent SDS-PAGE analysis. In order to allow different kinases to exert optimal activity during the assay, various buffer compositions were used. In the presence of Mg 2ϩ , a strong phosphorylation of a 55-kDa band was obtained ( Fig. 1, lanes 1 and 3-5). Replacement of Mg 2ϩ with Mn 2ϩ largely abolished this activity (Fig. 1, lane 2). Under conditions designed to stimulate protein kinase C (Fig. 1, lane  3), cAMP-dependent protein kinase (Fig. 1, lane 4), calcium/ calmodulin-dependent kinase (Fig. 1, lane 5), or cGMP-dependent protein kinase (not shown), phosphorylation of the 55-kDa band was not stimulated, and no other phosphorylated band of comparable intensity was detected, although some additional bands became evident. This study focused on the phosphorylation of the 55-kDa band under conditions similar to those given for Fig. 1, lane 1. Analysis of the time dependence of specific incorporation of radioactivity under conditions promoting high enzymatic turnover showed that the phosphorylation was substantial with regard to the total amount of isolated receptors (Fig. 2). The addition of 100 M sodium vanadate to the assay buffer had no effect, indicating the absence of a relevant phosphatase activity (not shown). No qualitative differences of the phosphorylation patterns were observed with concentrations of ATP ranging from 50 M to 400 M (results not shown).
The Endogenous Kinase Activity Preferentially Phosphorylates GABA A Receptor ␤3-Subunits-Following phosphorylation, GABA A receptor immunoprecipitates were enzymatically deglycosylated, subjected to SDS-PAGE, and blotted to nitrocellulose membranes, parallel to nondeglycosylated probes. Deglycosylation was done in order to unequivocally differentiate between receptor subunits and subunit isoforms and tubulin. It was indeed found that some proteins comigrating around 55 kDa (i.e. the GABA A receptor ␤2-subunit (Fig. 3, lane 7), the GABA A receptor ␤3-subunit (Fig. 3, lane 11), and ␣-tubulin (Fig. 3, lane 27)) showed a differential migration after deglycosylation (Fig. 3, lanes 8, 12, and 28). Radioactivity on the blot was detected by autoradiography, and the resulting pattern of labeled bands was compared with the signals obtained after immunodecoration with specific antibodies recognizing GABA A receptor ␣1-, ␤1-, ␤2-, ␤3-, ␤2/3-, and ␥2-subunits and ␣-tubulin (Fig. 3). After deglycosylation, the number of radioactive bands increased. This might be due to incomplete deglycosylation or to a partial protein degradation during deglycosylation. The predominant radioactive signals at 55 kDa in the nontreated probe (Fig. 3, lanes 1, 5, 9 , 13, 17, 21, and 25, upper band) and at 52 kDa in the deglycosylated probe (Fig. 3, lanes 2, 6, 10, 14, 18, 22, and 26, upper band) colocalized best with the signals obtained with an antibody specific for GABA A receptor ␤3subunits (Fig. 3, lanes 9 -12). Also, a weak radioactive band at 36 kDa, possibly representing a degradation product of ␤3subunits generated during deglycosylation by contaminating proteases, comigrated with a ␤3-subunit positive signal (Fig. 3,  lanes 14 and 16). Only one other antibody, recognizing GABA A receptor ␤2and ␤3-subunits, also reacted with the dominant radioactive bands of glycosylated and deglycosylated probes (Fig. 3, lanes 13-16), confirming the result with the ␤3-subunitspecific antibody. The additional bands recognized by the ␤2/ 3-subunit-specific antibody were due to the detection of ␤2subunits, as evident from the comparison with a ␤2-subunitspecific antibody (Fig. 3, lanes 7-8). With this antibody, as well as with the antibodies directed against GABA A receptor ␣1-, ␤1-, and ␥2-subunits and for ␣-tubulin, a match with each of the two predominant radioactive bands of the receptor probes (55 kDa for the glycosylated receptor, 52 kDa for the deglycosylated probe) was not obtained. These results indicated a preferential phosphorylation of GABA A receptor ␤3-subunits by the GABA A receptor-associated kinase activity. If it is assumed that ␤3-subunits are present in the receptor in a stoichiometry of 1:1 with ␣1-subunits, phosphorylation of ␤3-subunits almost occurs stoichiometrically (see Fig. 2). However, it should be pointed out that this probably represents an underestimate of the stoichiometry, since ␣1-subunits presumably assemble preferentially with ␤2and not with ␤3-subunits (9).
In 8 of 19 experiments, in addition to the dominant 55-kDa band also a minor 50-kDa band was observed to be phosphorylated (see Fig. 3, lane 1). The minor radioactive bands at 50 kDa in the nondeglycosylated and at 45 kDa in the deglycosylated probe could be matched with immunopositive signals for GABA A receptor ␣1-subunits (Fig. 3, lanes 17-20), ␤1-subunits (Fig. 3, lanes 1-4), and also ␤3-subunits (Fig. 3, lanes 9 -12). The immunostaining for the ␤1-subunit revealed a molecular mass of roughly 50 kDa for the glycosylated protein. This value is clearly lower than expected from the sequence similarity between the ␤1-subunit and the ␤2and ␤3-subunits, both migrating at 55 kDa. Most likely, this discrepancy can be explained by partial degradation of the ␤1-subunit during receptor purification, resulting in the loss of a terminal portion of the protein.

A Fusion Protein Containing an Intracellular Portion of the GABA A Receptor ␤3-Subunit Can Serve as Kinase Substrate-
The GABA A receptor immunoprecipitate was incubated, under conditions promoting high enzymatic turnover, with a fusion protein consisting of amino acid residues 345-408 of the intracellular loop of the GABA A receptor ␤3-subunit fused to MBP (MBP-␤3-(345-408)). A stoichiometry of phosphorylation of 0.79 mol of bound phosphate/mol of MBP-␤3-(345-408) was obtained (Fig. 4). This result indicated the usefulness of fusion proteins of intracellular loops of GABA A receptor subunits for further experiments. In another experiment under comparable conditions, a similar enzymatic activity was found.
The GABA A Receptor-associated Kinase Phosphorylates Serine Residues-The side chain specificity of the GABA A receptor-associated kinase activity was analyzed by two-dimensional thin layer chromatography of phosphoamino acids after acidolysis of the phosphorylated substrate proteins. A strict specificity of the kinase activity for serine residues was found both for the native ␤3-subunit of GABA A receptors (Fig. 5A) and for the MBP-␤3-(345-408) fusion protein (Fig. 5B). Phosphorylated native ␤3-subunits of GABA A receptors as well as for the MBP-␤3-(345-415) fusion protein were subjected to limited proteolysis by Lys-C (Fig. 5, C and D). Both phosphorylated entities resulted in each case in very similar phosphopeptide maps, indicating that the bovine ␤3-subunit results in identical phosphorylated proteolyzed fragments as the rat MBP-␤3-(345-415) fusion protein. The multiple radioactive spots could be indicative of either incomplete proteolysis or of multiple substrate sites.
These results demonstrated the phosphorylation of residues from intracellular loops of GABA A receptor ␤-subunits but not ␣1and ␥2-subunits by the receptor-associated kinase activity. While only the ␤3-subunit becomes phosphorylated in the native GABA A receptor (Fig. 3), intracellular loop constructs from all ␤-subunits investigated served as kinase substrate, with a preference for that from the ␤1-subunit.
Ser 408 of the ␤3-Subunit Is a Major Target for Phosphorylation-Several mutant MBP-␤3-(345-415) fusion proteins were prepared to identify the position of phosphorylated serine residues in the expressed portion of the intracellular loop of the GABA A receptor ␤3-subunit. The mutations S381A, S395A, and S407A did not reduce phosphate incorporation (Fig. 7B,  lanes 4 -6). The mutation S408A decreased the extent of phosphorylation strongly by 49 Ϯ 8% (n ϭ 3 experiments; see Fig.  7B, lane 7) and identified serine 408 as a major phosphorylation target. The fact that the neighboring serine residue 407 is not used as a substrate points to a highly selective phosphorylation reaction. This experiment was performed three times with comparable results.
The Kinase Activity Is Dissociated from the Receptor by Media of High Ionic Strength-In addition to the standard preparation of GABA A receptors derived from Triton X-100-solubilized membranes (Fig. 8A, lanes 1-3), an immunoprecipitate obtained from Zwittergent 3-14-solubilized membranes (32) was also investigated. This preparation results in a different GTAP composition (Fig. 8A, lanes 4 -6). With this preparation, MBP-␤3-(345-408) phosphorylation comparable with that of the standard receptor preparation was found (Fig. 8B, lanes 1  and 4). In addition, the phosphorylation of the MBP-␤3-(345-415; S408A) mutant was reduced to a similar degree (by 54%; mean of two experiments) as with the preparation from Triton X-100-solubilized membranes. Furthermore, the receptor-associated kinase phosphorylated a 55-kDa band of the GABA A receptor preparation, and no phosphate incorporation into the fusion proteins GST-␣1-(328 -382) and MBP-␥2-(319 -366) was found (data not shown). These results imply that both receptor preparations are associated with the same kinase activity.
Incubation of both GABA A receptor preparations with buffers of high ionic strength (high salt extraction) leads to the extraction of some of the associated proteins named GTAPs (32). A silver stain of the high salt-extracted receptor preparation from TX-100 is shown in Fig. 8A, lane 2, and preparation from Zwittergent 3-14 is shown in Fig. 8A, lane 5. The protein composition of the respective extracts is evident from Fig. 8A,  lanes 3 and 6. We compared the phosphorylation of MBP-␤3-(345-408) by nonextracted (Fig. 8B, lanes 1 and 4) and high salt-extracted immunoprecipitates (Fig. 8B, lanes 2 and 5) and salt extracts (Fig. 8B, lanes 3 and 6). In both detergents, high salt extraction strongly decreased kinase activities. Nevertheless, it was evident that the kinase activities remaining after high salt extraction were found to distribute almost entirely  8. Phosphorylation of MBP-␤3-(345-408) after high salt extraction of immunoprecipitates. GABA A receptor preparations obtained from Triton X-100- (lanes 1-3) or Zwittergent 3-14-(lanes 4 -6) solubilized membranes were treated with buffer 1 including 0.6 M NaCl. Nontreated beads (lanes 1 and 4), extracted beads (lanes 2 and 5), and the extracts (lanes 3 and 6) were analyzed by silver staining (A) and autoradiography of the phosphorylated fusion protein MBP-␤3-(345-408) (B).
into the salt extracts (Fig. 8B, lanes 3 and 6), with no activity phosphorylating MBP-␤3-(345-408) remaining in the immunobeads after treatment with high ionic strength (Fig. 8B, lanes 2  and 5). Comparable results were obtained in three similar experiments. The silver stain of the different preparations showed (Fig. 8A) that the only major protein present in the Zwittergent 3-14 extract, GTAP34, was also present in the Triton X-100 extract. This finding indicated that GTAP34 may be identical to the GABA A receptor-associated kinase. More evidence for this identity was provided by additional experiments described below (see Figs. 9 and 10).
The Kinase Activity Accepts Immunoprecipitated ␤2and ␤3-Subunits and to a Limited Extent Histone as a Substrate-The kinase activity was purified by immunoprecipitation in Zwittergent 3-14 and subsequent high salt treatment of the immunobeads to release the kinase activity into the supernatant (see Fig. 8). Bovine membranes were solubilized under denaturing conditions using SDS; subsequently, the SDS was quenched in Triton X-100, and GABA A receptor ␤2and ␤3subunits were immunoprecipitated using subunit-specific antibodies. In both cases, immunoprecipitated subunits were accepted as a substrate by the kinase (Fig. 9A). If the kinase was offered MBP-␤3-(345-415) fusion protein or histone as a substrate (Fig. 9B), histone was also used as a substrate, albeit to a smaller extent as compared with MBP-␤3-(345-415) fusion protein, which amounted to about 60% on a molar basis (two experiments). When histone was added directly to the immunoprecipitate and the combination was subjected to phosphorylation, histone was also accepted as a substrate, but it was phosphorylated to a much smaller extent than the native receptor subunit ␤3 (Fig. 9C). As stated above, ␤3 could only be estimated as a maximum amount. On a molar basis, it may be calculated that phosphorylation of histone amounts to less than 14% of phosphorylation of ␤3.
GTAP34 Comigrates with the MBP-␤3-(345-408)-phosphorylating Kinase Activity in Gel Filtration of High Salt Extract-High salt-extracted GTAPs were separated by gel filtration, and aliquots of collected fractions were silver-stained (Fig. 10A) or assessed for kinase activity using MBP-␤3-(345-408) as substrate (Fig. 10B). The kinase activity comigrated with GTAP34. This strongly indicates identity of GTAP34 with the GABA A receptor-associated serine kinase phosphorylating GABA A receptor ␤-subunits. Repetition of the experiment confirmed this result.
The prolonged exposure to high salt concentrations during gel filtration resulted in a strong decrease in kinase activity. The long exposure time needed for autoradiography (5 days compared with 1 h in Fig. 8) revealed the autophosphorylation of a protein with a molecular mass of approximately 57 kDa (Fig. 10C). Due to its low signal strength and because of overlap with the strong signals from GABA A receptor ␤-subunits and from the MBP-␤3-(345-408) fusion protein, this signal was not detected in the previous experiments. Although this signal may be due to the phosphorylation of a 57-kDa protein by trace amounts of GTAP34, it could also represent an autophosphorylating activity of another GABA A receptor-associated kinase. DISCUSSION We describe here a GABA A receptor-associated protein kinase with specificity for serine residues. Regulation of GABA A receptor function by direct phosphorylation through serine/ threonine kinases has been comprehensively studied (see Introduction). Biochemical features of the serine kinase described here are (a) the independence on cAMP and cGMP, (b) the independence on Ca 2ϩ and phospholipids, (c) the independence on Ca 2ϩ /calmodulin, and (d) the likely molecular mass of 34 kDa. These biochemical properties, including the molecular weight, distinguishes it from cAMP-dependent protein kinase (39 -41 kDa for the catalytic subunit), cGMP-dependent protein kinase (75-86 kDa), protein kinase C (Ͼ74 kDa for all the different isoforms), and Ca 2ϩ /calmodulin-dependent kinases (Ͼ37 kDa for all the different isoforms, all described in Ref. 47). All of these types of kinase have been shown to affect GABA A receptor function in different expression systems and neuronal preparations. Phosphorylation of the GABA A receptor by the present serine kinase may be regarded as a potential novel mechanism to exert cellular control on GABA A receptor func- tion. This kinase merits special interest because of its physical association with the receptor.
The GABA A receptor-associated kinase has a clear preference for ␤3-subunits in the intact receptor. ␣1and ␥2-subunits were not used as a kinase substrate either in the intact receptor or in overexpressed and purified loop constructs. Phosphorylation of ␤2-subunits was not evident, and that of ␤1-subunits, if occurring at all in native receptors, was quantitatively negligible. The results of previous studies on GABA A receptorphosphorylating activities associated with the receptor are substantially different from those of the study presented here. The serine kinase activities investigated by Sweetnam et al. (30) and Bureau and Laschet (31) showed strong phosphorylation of ␣-subunits, while phosphorylation of ␤-subunits was found to be either unlikely (30) or even absent (31). In addition, preferential kinase activation by Mn 2ϩ as compared with Mg 2ϩ was found (31), in contrary to the kinase described here.
MBP-loop fusion proteins of parts of the intracellular loops of all ␤-subunits isoforms could be phosphorylated, with a preference for ␤1. This contrasts with the observations made with native receptors and indicates, that the recombinant substrate proteins may not be not fully representative for the native kinase substrates. The bacterially overexpressed parts of the intracellular loops may have assumed a conformation different from that found in the native proteins. That this is indeed the case is made likely by the observations on the phosphorylation of different GABA A receptor ␤-subunits by cAMP-dependent protein kinase (25), where a similar apparent discrepancy was described. Alternative explanations seem to be less likely. For example, the weak or even absent phosphorylation of the native ␤1-subunit could possibly be due to a preferential degradation of ␤1-subunits as compared with ␤3-subunits during receptor purification, but the removal of a short N-or C-terminal portion from the ␤1-subunit would not be expected to have an effect on the ability of the ␤1-subunit to serve as a substrate. Also, an overrepresentation of the ␤3-subunit compared with ␤1and ␤2-subunits in the receptor preparation seems unlikely, since the ␣1-subunit, the subunit recognized and precipitated by the antibody bd24, is expected to be preferentially coassembled with ␤2but not with ␤3-subunits (9). The fact that the ␤3-subunit is probably a quantitatively minor protein of the receptor preparation underlines the specificity of the present serine kinase.
Roughly half of phosphate incorporation into a fusion protein containing intracellular sequences of the GABA A receptor ␤3subunit occurred at Ser 408 . That Ser 407 evidently was not used as a substrate pointed to a highly discriminating phosphorylation reaction. It is interesting to note that the homologous serine residue to ␤3 Ser 408 in the ␤1-subunit is a substrate for protein kinase C (18,19,23), for cAMP-dependent protein kinase (17,23), and for cGMP-dependent protein kinase and Ca 2ϩ /calmodulin-dependent kinase (27); and in the ␤2-subunit it is a substrate for protein kinase C (16). Phosphorylation of the corresponding residues resulted in alteration of GABA A receptor function (16 -19). From these findings, it can be predicted that GABA-evoked responses of ␤3-subunit-containing GABA A receptors can be modulated by the present serine kinase. In this respect, also the dual mode of GABA A receptor modulation by cAMP-dependent protein kinase-mediated phosphorylation recently described by McDonald et al. (25) is of interest. Phosphorylation of the ␤3-subunit at both serine residues Ser 408 and Ser 409 enhanced GABA A receptor function, while Ser 409 phosphorylation alone of a mutated ␤3-subunit was shown to inhibit GABA A receptor-mediated chloride currents. It is important to point out here that McDonald et al. (25) use a different numbering system of amino acid residues, our Ser 407 and Ser 408 corresponding to their Ser 408 and Ser 409 in ␤3-subunits. Future studies on the action of the kinase on functionally expressed GABA A receptors containing different ␤-subunits will be pivotal to precisely determine its functional consequences.
It is interesting to note that the protein kinase was obtained by immunoprecipitation using an antibody with specificity for the ␣1-subunit. Most probably, the kinase is associated directly with GABA A receptors containing an ␣1-subunit. However, the possibility cannot be excluded that the kinase is associated with another protein or with a GABA A receptor is associated itself directly or indirectly with an ␣1-subunit-containing GABA A receptor. In any case, the link between ␣1-subunitcontaining GABA A receptors and the kinase has to be resistant to overnight exposure to TX-100 and Zwittergent 3-14.
The strict specificity of the described protein kinase for ␤3subunits of the GABA A receptor seems to be lost once the kinase is released from the receptor into solution. At least the ␤2-subunit and, to a limited extent, histone (Fig. 9, A and B) can then additionally be used as a substrate. Histone is a very weak substrate for the bound kinase (Fig. 9C). It appears that the apposition of the protein kinase to the substrate plays an important role in its substrate specificity.
Several experiments indicate an identity of the present GABA A receptor-associated serine kinase with GTAP34, a protein that has recently been shown to be associated with GABA A receptors under the present conditions (32). First, the kinase and GTAP34 are present in the immunoprecipitates obtained from Triton X-100 and Zwittergent 3-14; second, extraction of the immunobeads with a medium containing high ionic strength eluted GTAP34 as well as the serine kinase; third, in gel filtration experiments of high salt extract, GTAP34 and the kinase activity co-eluted in the same fractions. This identification on the protein level should lead to a description at the DNA level.
In conclusion, the data presented here point to GTAP34 as being a GABA A receptor-associated serine kinase with preference for ␤3-subunits as a substrate. Study of this kinase may contribute to the further understanding of the physiological significance of GABA A receptor subunit diversity (25, 48 -50). In addition, GTAP34 could serve as a target for drugs specifically modulating the function of ␤3-subunit-containing GABA A receptors.