Activation of Heterotrimeric G Proteins by a High Energy Phosphate Transfer via Nucleoside Diphosphate Kinase (NDPK) B and Gβ Subunits

G protein βγ dimers can be phosphorylated in membranes from various tissues by GTP at a histidine residue in the β subunit. The phosphate is high energetic and can be transferred onto GDP leading to formation of GTP. Purified Gβγ dimers do not display autophosphorylation, indicating the involvement of a separate protein kinase. We therefore enriched the Gβ-phosphorylating activity present in preparations of the retinal G protein transducin and in partially purified Gi/oproteins from bovine brain. Immunoblots, autophosphorylation, and enzymatic activity measurements demonstrated enriched nucleoside diphosphate kinase (NDPK) B in both preparations, together with residual Gβγ dimers. In the retinal NDPK B-enriched fractions, a Gβ-specific antiserum co-precipitated phosphorylated NDPK B, and an antiserum against the human NDPK co-precipitated phosphorylated Gβγ. In addition, the NDPK-containing fractions from bovine brain reconstituted the phosphorylation of purified Gβγ. For identification of the phosphorylated histidine residue, bovine brain Gβγ and Gtβγ were thiophosphorylated with guanosine 5′-O-(3-[35S]thio)triphosphate, followed by digestion with endoproteinase Glu-C and trypsin, separation of the resulting peptides by gel electrophoresis and high pressure liquid chromatography, respectively, and sequencing of the radioactive peptides. The sequence information produced by both methods identified specific labeled fragments of bovine Gβ1 that overlapped in the heptapeptide, Leu-Met-Thr-Tyr-Ser-His-Asp (amino acids 261–267). We conclude that NDPK B forms complexes with Gβγ dimers and contributes to G protein activation by increasing the high energetic phosphate transfer onto GDP via intermediately phosphorylated His-266 in Gβ1 subunits.

Heterotrimeric G proteins play a pivotal role in many signal transduction pathways in eukaryotic cells. They consist of a guanine nucleotide-binding ␣ subunit (40 -52 kDa), a ␤ subunit (33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43), and a ␥ subunit (6 -10 kDa). The latter two act as a functional unit and only dissociate upon denaturation. Both G␣ and G␤␥ are required for receptor-induced G protein activation and can trigger effector functions (for reviews see Refs. [1][2][3]. Heterotrimeric G proteins are activated by a GDP/GTP exchange catalyzed by G protein-coupled receptors. Furthermore, we and other laboratories provided evidence that phosphotransfer reactions can participate in G protein activation in vitro by formation of GTP. There is amble evidence that nucleoside diphosphate kinase (NDPK) 1 contributes to G protein activation by replenishment of GTP from ATP and GDP (for reviews see Refs. 4 -6). Hypotheses suggesting a direct in situ phosphorylation of GDP bound to G␣ and monomeric G proteins (7)(8)(9) are most likely based on artifacts. Also complex formation of NDPK with G proteins and channeling of NDPKformed GTP into G␣ (10,11) have not yet been proven beyond doubt.
A phosphotransfer reaction that uses G␤ subunits as phosphorylated intermediates has been observed in various tissues (12)(13)(14)(15)(16). In this reaction, the ␥-(thio)phosphate group of GTP or its analog GTP␥S is transferred onto a histidine residue of G␤. Apparently, a membrane-bound, so far unknown co-factor is required to achieve this phosphorylation (14,15). Out of the labile high energy phosphoamidate bond, the phosphate can be retransferred onto GDP to form GTP, which then can activates G i and G s proteins and thus regulates, for example, adenylyl cyclase activity (17). Nevertheless, the exact significance of this phosphotransfer reaction remains elusive.
The purpose of this study was to test the hypothesis that NDPK may represent the unknown co-factor and contributes to the phosphate transfer via G␤. We report here that by the attempts to purify the co-factor from the retinal G protein transducin (G t ) or bovine brain membranes, we specifically enriched the NDPK B isoform. We will further provide evidence for a complex formation of G␤␥ with NDPK B and for a specific phosphorylation of His-266 of G␤ 1 in this complex.

Preparation of Rod Outer Segment Membranes, Purification of NDPK, Purification of G t and its Subunits, and Resolution of the G␤-
Phosphorylating Activity from G t -Rod outer segment (ROS) membranes were prepared from illuminated bovine retina according to * This work was supported by grants from the Deutsche Forschungsgemeinschaft and the Fonds der Chemischen Industrie. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
** To whom correspondence should be addressed. Tel.: 49-621-330030; Fax: 49-621-3300333; E-mail: thomas.wieland@urz.uni-heidelberg.de. Papermaster and Dreyer (18). NDPK was purified to homogeneity from the cytosol essentially as described (19). G t was eluted from the membranes by repeated washing with the GTP analog, GppNHp (100 M), as described (12). The G t -containing eluate was concentrated by pressure filtration with a PM 10 membrane (Amicon, Witten, Germany), and unbound GppNHp was removed by a gel PD-10 gel filtration column (Amersham Biosciences) (12). About 2 mg of purified G t was applied onto a Blue-Sepharose CL6B column (10-ml bed volume) at a flow rate of 0.5 ml/min and separated into G t ␤␥ and G t ␣ as described (12).
The G t ␣-containing fractions were pooled and applied onto a 5-ml hydroxyapatite column (Econo Cartridge CHTII; Bio-Rad) at a flow rate of 0.3 ml/min. The column was washed with 20 ml of a buffer containing 10 mM Tris-HCl, pH 6.5, 6 mM MgCl 2 , 1 mM dithiothreitol, and 20% (v/v) glycerol. Bound proteins were eluted with a KH 2 PO 4 gradient (0 -400 mM). Fractions of 1 ml were collected and analyzed for the content of G t subunits, G␤-phosphorylating activity, and NDPK activity. Fractions containing the G␤-phosphorylating activity were pooled and stored at Ϫ80°C.
Purification of G i/o Proteins, G␤␥, and the G␤-Phosphorylating Activity from Bovine Brain Membranes-Membranes from bovine brain were prepared as described (20). Proteins were solubilized from membranes (4 g of protein) by stirring for 1 h at 4°C in 800 ml of TEM buffer (20 mM Tris-HCl, pH 8, 1 mM EDTA, 20 mM 2-mercaptoethanol) containing 1% (w/v) 1-octyl-1-␤-D-thioglucopyranoside (Biomol, Hamburg, Germany). After centrifugation for 40 min at 110,000 ϫ g, ethylene glycol was added to the supernatants to a final concentration of 30%. 900 ml of this extract, i.e. ϳ800 mg of protein, were used for three subsequent steps of liquid chromatography, carried out at 4°C, using a fast protein liquid chromatography device. Elution was monitored by continuous measurement of absorbance at 280 nm. The extract was first loaded onto a column (10 ϫ 13 cm) containing 1 liter of DEAE-Sepharose (Amersham Biosciences) at a flow rate of 0.5 ml/min. After washing with TEMEC (TEM buffer containing additionally 30% ethylene glycol and 0.9% (w/v) sodium cholate), elution was performed with a linear gradient of NaCl (0 -800 mM, volume 2.2 liters). Fractions of 12 ml were collected. G i/o proteins eluted from the column at about 450 mM NaCl were further purified and separated into their subunits as described (20,21). Fractions capable of phosphorylating G␤ subunits were identified in the phosphorylation assay (see below), using 0.3 g of bovine brain G␤␥ as substrate. Positive fractions eluting at 220 -270 mM NaCl were pooled and concentrated by pressure filtration using a PM 10 membrane. This pool (about 35 mg of protein) was then loaded onto a column (3 ϫ 14 cm) containing 100 ml of hydroxyapatite (E. Merck, Darmstadt, Germany) at a flow rate of 0.3 ml/min. The column was washed with TEMEC. Proteins were eluted with a linear gradient of K 2 HPO 4 /KH 2 PO 4 (0 -400 mM, pH 8, volume 300 ml), collecting fractions of 5 ml. Phosphorylation-positive fractions eluting at phosphate concentrations of 270 -290 mM were pooled and concentrated, and the buffer was exchanged into HEMEC (20 mM HEPES-NaOH, pH 6.5, 1 mM EDTA, 20 mM 2-mercaptoethanol, 30% ethylene glycol, 0.9% sodium cholate). This pool (about 6 mg of protein) was then loaded onto a XK16/20 column (Amersham Biosciences) containing 25 ml of EMD SO 3 650(S) (E. Merck), at a flow rate of 0.2 ml/min. The column was washed with HEMEC. Elution from the cation exchange column was performed with a linear gradient of NaCl (0 -1 M, volume 80 ml) in HEMEC. Fractions of 2 ml were collected. Positive fractions eluting between 350 and 400 mM NaCl were concentrated by centrifugation in a Centricon-30 device (Amicon). Chromatography media were regenerated according to the respective manufacturer's recommendations.
Phosphorylation and Thiophosphorylation of G␤ Subunits and NDPK-The indicated amounts of bovine brain or transducin G␤␥ dimers and column fractions of the purification procedures were phosphorylated with 10 nM [␥-32 P]GTP (PerkinElmer Life Sciences) for the indicated periods of time at 30°C in a reaction buffer containing 50 mM triethanolamine hydrochloride, pH 7.4, 150 mM NaCl, 2 mM MgCl 2 , 1 mM EDTA, and 1 mM dithiothreitol in a total volume of 20 l. For thiophosphorylation of G␤, 20 nM [ 35 S]GTP␥S was used, and incubation was for 30 min. The reaction was terminated by the addition of one volume of sample buffer, followed by 1 h of incubation at room temperature. Proteins were separated by discontinuous SDS-PAGE on gels containing 10 -12% (w/v) acrylamide and autoradiographed.
Determination of NDPK Activity-The determination of the NDPK activity in preparations obtained from solubilized bovine brain membranes or G t was performed in a reaction mixture (50 l total volume) containing 50 mM triethanolamine hydrochloride, pH 7.4, 2 mM MgCl 2 , 50 M ATP, and 1 M [8-3 H]GDP (Amersham Biosciences). The reaction was started by addition of 20 l of the protein-containing solutions and conducted for 30 min at room temperature. Termination of the reaction and analysis of GTP formation were performed essentially as described (22).
Immunoprecipitation-The pooled fractions (60 l) from the hydroxyapatite column, containing the peak activity of NDPK from G t , were phosphorylated in a final volume of 80 l at 30°C for 7 min. The reaction was terminated by placing the mixture on ice and adding an equal volume of 50 mM EDTA, pH 7.4. Precipitation buffer (280 l), containing 150 mM NaCl, 10 mM Tris-HCl, pH 7.4, 25 mM EDTA, 1 mM dithiothreitol, 1 mM NaF, and 0.2 M phenylmethylsulfonyl fluoride, was added, followed by the addition of 10 l of the same buffer containing 1 mg of protein A-Sepharose beads. After incubation on ice for 20 min and centrifugation (26,000 ϫ g, 15 min), the clear supernatant was taken and supplemented with 0.6 g of anti-G␤ (T-20) or 4 g of anti-NDPK (C-20) antiserum and 5 mg of protein A-Sepharose beads. The mixture was gently shaken for 2 h at 4°C. Protein A-Sepharose beads were pelleted and washed three times with precipitation buffer containing 300 mM NaCl. Bound proteins were eluted from the protein A-Sepharose beads by adding 50 l of SDS buffer. The samples were kept at room temperature for at least 1 h before loading onto a 10% polyacrylamide gel.
Treatment of G t ␤␥ with Diethyl Pyrocarbonate-Purified G t ␤␥ (30 g) was incubated for 10 min at room temperature in a buffer (100 l) containing 10 mM triethanolamine hydrochloride, pH 7.4, and 10 mM diethyl pyrocarbonate. The reaction was terminated by the addition of 25 mM EDTA. Free diethyl pyrocarbonate and EDTA were subsequently removed by repeated buffer exchange into 20 mM Tris-HCl, pH 8, 1 mM EDTA, 20 mM 2-mercaptoethanol using a Microcon-10 device (Amicon).
Endoproteinase Glu-C Digest of Thiophosphorylated G␤␥, Purification, and Sequencing of Peptides-Bovine brain G␤␥ (15 g of protein) was thiophosphorylated with 100 nM [ 35 S]GTP␥S and 1 g of NDPK B-enriched co-factor as described above, in a total volume of 100 l. The reaction was terminated by addition of 50 l of 3-fold concentrated SDS-PAGE sample buffer and 10 g of endoproteinase Glu-C (Roche Molecular Biochemicals). Proteins were digested for 3 h at 37°C, and the proteolytic fragments were separated on a Tris-Tricine-based system (23). After electrotransfer onto a nitrocellulose membrane, the labeled fragment was identified by autoradiography and excised. The peptide was eluted, and its amino acid sequence was determined by automated Edman degradation in an Applied Biosystems, Inc. (Foster City, CA) model 476A protein sequencer.
Tryptic Digest of Thiophosphorylated G t ␤␥, Resolution, and Sequencing of Peptides-Unmodified G t ␤␥ (5 g) was thiophosphorylated with 100 nM [ 35 S]GTP␥S and 1 g of co-factor as described above for 30 min in a total volume of 40 l. Thereafter, 10 g (20 l) of unlabeled thiophosphorylated G t ␤␥ (12), 35 l of 50 mM NH 4 CO 3 , pH 7.7, 5 l acetonitrile, and 2.5 g of trypsin were added. After incubation overnight at 37°C, the digest was injected directly into an analytical (2 ϫ 250 mm) Ultrasep Es 100 (E. Merck, Darmstadt, Germany) C 18 reverse phase HPLC column, equilibrated with a mixture of 90% Solution A (0.2% hexafluoroacetone, NH 3 , pH 8.6) and 10% Solution B (0.02% hexafluoroacetone, 84% methylcyanide). The gradient was run from 10 to 80% Solution B at an increase rate of 1%/min with a flow rate of 80 l/min. The absorbance was monitored at 215 and 295 nm, and peak fractions were collected. 35 S-Labeled peptides were detected by liquid scintillation counting of 3 l of each fraction. The sequence of the peptide in the fraction containing the radioactivity peak was determined by automated Edman degradation.

RESULTS
G␤-Phosphorylating Activity in G t and Complex Formation of G t ␤␥ with NDPK-The ␤ subunits of the retinal G protein G t can be transiently (thio)phosphorylated by GTP or GTP␥S and an enzymatic activity present in ROS membranes (12). More recently, it was demonstrated that soluble preparations of G t contain NDPK activity and that GTP phosphorylates a 36-kDa protein, most likely G␤ (22). We therefore attempted to identify the putative co-factor present in the G t preparations and investigated whether NDPK is involved in this reaction. For this purpose, G t was eluted from ROS membranes with 100 M of the stable GTP analog, GppNHp, which does not modify G t ␤␥ (12). G t ␤␥ and G t ␣ can be resolved from each other by affinity chromatography on Blue-Sepharose (12,22). Whereas G t ␤␥ has little affinity to the matrix, elution of G t ␣ requires high salt concentrations (500 mM KCl). Single protein bands at the apparent molecular weights of G t ␤␥ and G t ␣ were detected by SDS-PAGE and Coomassie Blue staining (not shown) of the fractions from the first and second peak, respectively. In search for the G t ␤-phosphorylating activity, the fractions were subjected to phosphorylation with [␥-32 P]GTP. In the G t ␣-containing fraction, two phosphorylated proteins (M m 18 and 36 kDa; see Fig. 1) were detected. Measurement of NDPK activity revealed a formation of 0.2 nmol GTP per mg of protein per min in this fraction. In line with data reported previously (12,22), no phosphorylated proteins and no NDPK activity was detected in the G t ␤␥-containing fractions. To separate the two phosphoproteins from G t ␣, ϳ1 mg of protein from the second peak was applied onto a hydroxyapatite column, and proteins were eluted with increasing concentrations of potassium phosphate (0 -400 mM). G t ␣ eluted from the column in a single peak at about 10 mM phosphate (fractions 23-26; see Fig. 1A). No NDPK activity and no phosphorylation by [␥-32 P]GTP were detected in the G t ␣ peak. The NDPK activity eluted from the column at ϳ250 mM phosphate (Fig. 1B). As shown in the inset of Fig. 1B, this fraction contained both the 18-and 36-kDa proteins phosphorylated by [␥-32 P]GTP. Immunoblot analysis showed that this fraction contained NDPK protein, G t ␤␥, and traces of G t ␣.
To identify the two proteins phosphorylated by [␥-32 P]GTP, we immunoprecipitated G␤ and NDPK with specific antisera after phosphorylation. Identical amounts of IgG were used as controls. As shown in Fig. 2, the NDPK-specific antiserum precipitated NDPK as a doublet migrating at 18 -20 kDa and an additional phosphoprotein at 36 kDa. The G␤-specific antiserum precipitated phosphorylated G␤ (36 kDa) and an additional phosphoprotein at 18 -20 kDa. Some minor phosphorylated bands in the range of 28 to 30 kDa most likely represent degradation products of phosphorylated G␤.
Enrichment of the G␤-Phosphorylating Activity from Bovine Brain Membranes and Reconstitution of G␤ Phosphorylation-As G t differs from other G proteins by its solubility without detergent, we attempted to purify the G␤-phosphorylating activity from another tissue. Because heterotrimeric G proteins are abundant in bovine brain membranes, and the extent of G␤ phosphorylation is rather high compared with membranes of other tissues available in sufficient amounts (15), we first passed a detergent extract from bovine brain membranes over a DEAE column. G proteins that elute from the column at about 450 mM NaCl could no longer be phosphorylated with [␥-32 P]GTP (15) (not shown). Apparently, the co-factor that promotes phosphorylation had been separated from the majority of G proteins. Addition of the fraction eluting at about 250 mM NaCl to the G protein-containing fraction or purified bovine brain G␤␥ reconstituted the phosphorylation. The fraction that promoted G␤ phosphorylation was further purified by hydroxyapatite chromatography, followed by cation exchange chromatography using an EMD SO 3 650(S) column. After each step, positive fractions were identified by their potential to phosphorylate G␤. Equal amounts of protein (1 g) were used after each step to reconstitute the phosphorylation of purified bovine brain G␤␥ (0.5 g). Phosphorylated G␤ subunits were excised from SDS-PAGE gels, and the amount of radioactivity was detected by liquid scintillation counting, to estimate the specific activity after each purification step. The data are summa- FIG. 1. Resolution of NDPK from G t ␣ by chromatography on hydroxyapatite. A, 1 mg of the G t ␣ pool was applied onto an Econo Cartridge CHTII column. Bound proteins were eluted with the indicated potassium phosphate gradient (dotted line), and the protein content was monitored by absorbance at 280 nm. G t ␣ (inset) eluted from the column in a single peak at 10 mM phosphate. B, the NDPK activity of the eluate fractions was measured as formation of [ 3 H]GTP as described under "Experimental Procedures." The fractions containing the NDPK activity peak were pooled and phosphorylated with [␥-32 P]GTP. An autoradiograph after SDS-PAGE is shown (P, inset). The content of NDPK protein, G t ␤␥, and G t ␣ was determined by Western blot (WB, inset).  Table I. A small decrease in specific activity after the EMD SO 3 cation exchange column was noted. However, the chromatogram showed that 97% of total protein could be removed during this step. Thus, the decrease in specific activity might have been because of decaying enzymatic activity in the diluted protein fraction during longer storage at 4°C. SDS-PAGE and silver staining revealed that the fractions contained four to five faint protein bands in the molecular range from 20 to 50 kDa (not shown). Similar to the G␤-phosphorylating activity prepared from G t , NDPK activity (1.8 nmol of GTP formed per mg protein per min) and NDPK protein (immunoblot, autophosphorylation) could be detected (Fig. 3A). To test whether this preparation is able to reconstitute the phosphorylation of bovine G␤␥, increasing amounts of the EMD SO 3 650(S) pool (0.04 -0.4 g of protein) were phosphorylated with [␥-32 P]GTP in the absence and presence of purified G␤␥ (0.5 g) for 5 min at 30°C (Fig. 3B). Conversely, the phosphorylation of increasing amounts of purified bovine brain G␤␥ dimers promoted by a constant quantity of partially purified factor (200 ng of protein) was studied (Fig. 3C). The phosphorylation of G␤ increased with the amount of added co-factor fraction or G␤␥ dimer. Maximal phosphorylation was observed with 1-3 g of G␤␥ dimer. Similar data were obtained when G t ␤␥ was used as substrate (Fig. 3D). Previous results showed that pretreatment of membranes with diethyl pyrocarbonate prevents the phosphorylation of G␤ because of the ethoxycarbonylation of histidine residues (13,24). Similarly, G t ␤␥ pretreated with diethyl pyrocarbonate (10 mM) prior to phosphorylation with the cofactor-containing fraction exhibited no increase in phosphorylation of the ethoxycarbonylated G t ␤ (Fig. 3D). In addition, phosphorylated G␤ was sensitive to treatment with hydroxylamine (data not shown) that cleaves phosphohistidine (25). As shown in Fig. 3E, the co-factor-containing fraction also reconstituted the thiophosphorylation of G␤ by [ 35 S]GTP␥S. When higher amounts of the co-factor pool were used, the phosphorylated G␤ was detected also in the absence of added G␤␥ (Fig.  3, A, B, and D). Indeed, small amounts of G␤ could be identified in the co-factor pool by Western blot analysis (Fig. 3A).
Identification of the NDPK B Isoform in the G␤-Phosphorylating Co-factor Pools-Two major isoforms of NDPK, NDPK A (nm23-H1) and NDPK B (nm23-H2), are present in the bovine FIG. 3. Reconstitution of the G␤ (thio)phosphorylation of G␤␥. A, the co-factor required for G␤ phosphorylation in solubilized bovine brain membranes was enriched by subsequent anion exchange, hydroxyapatite, and cation exchange chromatography as described under "Experimental Procedures." The fractions obtained from the cation exchange column at 350 -380 mM NaCl were pooled. After phosphorylation with [␥-32 P]GTP for 5 min at 30°C, 800 ng of protein were subjected to SDS-PAGE. Phosphorylated proteins were visualized by autoradiography. G␤ and NDPK were detected by Western blotting with specific antisera. B, increasing amounts (40 -400 ng of protein) of the co-factor pool (CF) were phosphorylated in the absence (Ϫ) and presence (ϩ) of 0.5 g of purified bovine brain G␤␥ for 5 min at 30°C. C, increasing amounts of G␤␥ (10 -3000 ng) were phosphorylated in the presence of 200 ng of the cation exchange fraction. D, G t ␤␥ was treated with 10 mM diethyl pyrocarbonate as described under "Experimental Procedures." Thereafter, phosphorylation of G␤ by [␥-32 P]GTP was determined with the cation exchange fraction in the absence (Control) and presence of 2 g untreated (G t ␤␥) or diethyl pyrocarbonate-treated G t ␤␥ (G t ␤␥/DEPC). E, 200 ng of the cation exchange fraction were thiophosphorylated with 20 nM [ 35 S]GTP␥S in the absence (Control) and presence of 1 g of purified bovine brain G␤␥ for 30 min at 30°C. Autoradiographs after SDS-PAGE are shown.  (19) and other tissues (5,6). We therefore investigated whether one or both forms had been enriched during the purification procedures for the G␤-phosphorylating activity by Western blot analysis with subtype-specific antisera (kindly provided by Dr. Ioan Lascu, Bordeaux, France) (Fig. 4). Purified recombinant human NDPK A and B (also kindly provided by Dr. Ioan Lascu) and purified NDPK from the cytosol of bovine ROS served as positive controls. Although the anti-NDPK A antiserum was more sensitive than the anti-NDPK B antiserum, only NDPK B was detected in the G␤-phosphorylating fractions obtained from G t and bovine brain membranes.
Endoproteinase Glu-C and Tryptic Digest of Thiophosphorylated G␤␥ and Identification of the Phosphorylated Histidine in G␤ 1 -Endoproteinase Glu-C cleaves proteins at the C-terminal end of glutamate residues and can additionally be used for in gel digests of proteins in a SDS-containing buffer. Complete digest of G␤ 1 by endoproteinase Glu-C would produce several peptides in the range of 1000 to 13,500 kDa. Each contains not more than three histidine residues. We therefore asked whether we could detect a labeled phosphopeptide after in gel digest with endoproteinase Glu-C and separation of the peptides by high resolution SDS-PAGE (19). As shown in Fig.  5A, endoproteinase Glu-C digest of phosphorylated bovine brain G␤␥ produced a single labeled peptide (M m ϳ9 kDa). The same peptide is recognized by the anti-G␤ (T-20) antibody (Fig.  5B). To identify the phosphopeptide, 15 g of bovine brain G␤␥ were thiophosphorylated with 100 nM [ 35 S]GTP␥S and digested with endoproteinase Glu-C. After SDS-PAGE and electrotransfer onto nitrocellulose, the labeled peptide was identified by autoradiography and excised from the blot. The peptide was eluted and sequenced by Edman degradation (26). The experiment was repeated twice. The first run produced the sequence LMXYXXDXII. In the second run, the 14-amino acid sequence, LMTYSHDNIICGIT, could be identified. This sequence corresponds to the N terminus of a peptide (amino acid 261-340; M m 8657.71) that results from endoproteinase Glu-C digest of G␤ 1 and contains two histidine residues, His-266 and His-311.
Although G␤ 1 complexed with different G␥ subunits is abundant in bovine brain G␤␥ preparations, at least G␤ 2 ␥ x dimers are similarly present (27). We therefore used the well defined G t ␤␥, i.e. G␤ 1 ␥ 1 , for further analysis. Unmodified G t ␤␥ was thiophosphorylated with 100 nM [ 35 S]GTP␥S and the co-factor from bovine brain (see Fig. 3) and added to 10 g of unlabeled thiophosphorylated G t ␤␥. The mixture was digested overnight with trypsin, the peptides were separated by HPLC on a C 18 reverse phase HPLC column, and peak fractions were collected (Fig. 6A). An aliquot of each fraction (3 l) was counted for radioactivity. As shown in Fig. 6B, the labeled peptide eluted from the column in a sharp peak at fraction 22. The analysis of this fraction by Edman degradation revealed the sequence ADQELMXYSHD. It corresponds to the N terminus of an ex-

DISCUSSION
Meanwhile, several laboratories have reported the intermediate formation of a high energy phosphoamidate bond on a histidine residue of G␤ subunits (12)(13)(14)(15)(16)(17) that enables the formation of GTP by transferring this phosphate onto GDP. The phosphorylation of G␤ requires a co-factor that acts as a histidine kinase (15,16,28). In addition, early studies on an NDPKlike activity leading to formation of GTP from ATP and GDP in retinal extracts from frogs associated this so called GDP kinase activity with the G t ␤␥ dimer (29). In this study, we obtained several lines of evidence for a complex formation of the NDPK B isoform with G␤␥ dimers. First, NDPK B and G␤␥ were co-purified from both transducin and solubilized bovine brain membranes with different purification protocols. Apparently, the biochemical properties of the NDPK determined the purification of the complex. In contrast to G␤␥, NDPK binds tightly to Blue-Sepharose, and this property can be used to purify NDPK (19). Recently, the NDPK of Drosophila melanogaster was purified by a combination of anion exchange and hydroxyapatite chromatography (30). G␤ 1 ␥ 2 and G t ␤␥ exhibit only a weak interaction with hydroxyapatite and elute from the matrix at much lower concentrations of phosphate (31,32). Second, a G␤-specific antiserum co-immunoprecipitated NDPK, and conversely, an NDPK-specific antiserum co-immunoprecipitated G␤ from the retinal preparation. In agreement with our data more indirect evidences for an interaction of NDPK B with G t have been described previously by others (11,22). The presence of NDPK B and G␤␥ or heterotrimeric G proteins, however, is apparently not sufficient to obtain the complex formation between G␤␥ and NDPK B to reconstitute the phosphorylation of G␤. Most likely, another protein is required as scaffold for that complex. Recently, it was demonstrated that NDPK forms a complex with phocein, Eps 15, and dynamin I, a GTPase that plays a critical role in endocytosis (33). Within this complex, NDPK interacts with dynamin I through a proline-rich domain, whereas its interaction with phocein is ill defined. The data, however, indicate that complexes of NDPK with multiple proteins occur as mechanisms for local GTP replenishment within cells.
The question arises whether NDPK B within the complex is the histidine kinase required for the phosphorylation of G␤. Protein kinase activity for NDPK has been described. Serine and threonine phosphorylation by NDPK was found on histone 2b, casein, and ovalbumin (34,35). Moreover, NDPK phosphorylates the catalytic histidine residue in ATP citrate lyase (36), and other reports (37)(38)(39) indicate that NDPK is the phosphate donor for the phosphorylation of a histidine in annexin I. Most recently, Kowluru (28) reported that the phosphorylation of G␤ and histone 4 was largely increased by mastoparan, a known activator of NDPK and G proteins (40,41). In accordance with the presumed histidine kinase activity of NDPK B, an increase in G␤ phosphorylation was observed in membranes of H10 cells overexpressing wild-type NDPK B but not its catalytically inactive mutant H118N (42).
The NDPK B-enriched fraction obtained from bovine brain reconstituted the phosphorylation and thiophosphorylation of G␤ by GTP and GTP␥S, respectively. As observed before in membranes (13,15,16), phosphorylated G␤ was sensitive to hydroxylamine cleavage (25) and was prevented by ethoxycarbonylation with diethyl pyrocarbonate (23). Thus, all our results in the reconstituted systems are in agreement with earlier results obtained on the (thio)phosphorylation of a histidine in G␤. The data obtained from proteolytic digest of the thiophosphorylated bovine brain G␤␥ and G t ␤␥ revealed that His-266 of G␤ 1 is the phosphorylated residue in G␤. As shown in Fig. 7, the imidazolyl side chain of His-266 is freely accessible on the surface of the heterotrimeric G protein and can therefore be the target of protein phosphorylation. The seven other histidines are part of the G␤ propeller structure and thus are unlikely accessible to kinases. His-266 is conserved in G␤ 2 , G␤ 3 , and G␤ 4 but is found not in G␤ 5 , which has a lysine at the analogous position (43). Although we have no proof for this hypothesis at this time, we propose that, based on the high degree of homology mammalian, G␤ 1-G␤ 4 can be phosphorylated at His-266. The mainly neuronally expressed G␤ 5 functionally differs from the other G␤ subunits. In accordance with the lack of the respective histidine, we were not able to detect the phosphorylation of G␤ 5␥2 (kindly provided by Dr. B. Nü rnberg, Dü sseldorf, Germany) in our reconstitution assay (data not shown). Two non-mammalian G␤ subunits are also lacking a histidine at the analogous position. These are G␤ 2 of D. melanogaster, which has a proline at this position, and G␤ (Ste4p) from Saccharomyces cerevisiae. Interestingly, Ste4p has a 41-amino acid insertion at this position. Amino acids within this insertion are target to phosphorylation, which is necessary for adaptation of the mating response in yeast (44,45). In summary, His-266 is apparently not required for forma- tion of the seven propeller blade G␤ structure and is placed on the surface of the heterotrimer at a position where phosphorylation is likely to occur. Therefore, we propose that NDPK B phosphorylates the structurally exposed His-266 in G␤ in a stoichiometric complex with G␤␥.
A phosphotransfer from NTP to His-118 in NDPK B and subsequently onto His-266 of G␤ and further onto GDP might offer an explanation for several reports indicating a higher potency of NDPK-formed GTP compared with exogenously added GTP in G protein activation (10,46,47). However, the three-dimensional structure of the heterotrimeric G protein (48) (Fig. 7) indicates that His-266 is distant from the GDP molecule within the G␣ subunit. Thus, a direct transfer onto G␣-bound GDP is not supported by these structural data. Nevertheless, the drastic increase in adenylyl cyclase activity as a result of an overexpression of NDPK B and G s ␣ (42) might be an indication for the high efficiency of the phosphotransfer to activate G proteins.
Both purification procedures revealed that the vast majority of G proteins do not co-purify and thus are not complexed with NDPK B. Therefore, they are most likely not accessible to the phosphate transfer reaction. Although it is likely that purification, especially solubilization from the membrane environment, could cause a substantial dissociation of NDPK B from G proteins, the so far unknown scaffolding protein mentioned above could be the limiting factor within the complex of NDPK B and G␤␥. Moreover, a low abundance of the NDPK-G protein complex would fit into a concept where NDPK exclusively regulates the basal tone of G protein activities and where the vast majority of G proteins serve the "classical" receptor signal transduction, as discussed in the accompanying paper (42).