Interaction of Heterotrimeric G Protein Gαowith Purkinje Cell Protein-2

The heterotrimeric G protein Gαo is ubiquitously expressed throughout the central nervous system, but many of its functions remain to be defined. To search for novel proteins that interact with Gαo, a mouse brain library was screened using the yeast two-hybrid interaction system. Pcp2 (Purkinje cellprotein-2) was identified as a partner for Gαo in this system. Pcp2 is expressed in cerebellar Purkinje cells and retinal bipolar neurons, two locations where Gαo is also expressed. Pcp2 was first identified as a candidate gene to explain Purkinje cell degeneration in pcdmice (Nordquist, D. T., Kozak, C. A., and Orr, H. T. (1988) J. Neurosci. 8, 4780–4789), but its function remains unknown as Pcp2 knockout mice are normal (Mohn, A. R., Feddersen, R. M., Nguyen, M. S., and Koller, B. H. (1997) Mol. Cell. Neurosci. 9, 63–76). Gαoand Pcp2 binding was confirmed in vitro using glutathioneS-transferase-Pcp2 fusion proteins and in vitrotranslated [35S]methionine-labeled Gαo. In addition, when Gαo and Pcp2 were cotransfected into COS cells, Gαo was detected in immunoprecipitates of Pcp2. To determine whether Pcp2 could modulate Gαo function, kinetic constants k cat andk off of bovine brain Gαo were determined in the presence and absence of Pcp2. Pcp2 stimulates GDP release from Gαo more than 5-fold without affectingk cat. These findings define a novel nucleotide exchange function for Pcp2 and suggest that the interaction between Pcp2 and Gαo is important to Purkinje cell function.

receptor. Agonist-liganded receptors activate G␣ by inducing a change in conformation that leads to GDP release and GTPbinding. GTP-liganded G␣ dissociates from G␤␥, and both subunits can interact with a variety of intracellular effectors. G␣ and G␤␥ remain activated until the intrinsic GTPase activity of G␣ hydrolyzes GTP to GDP. The mechanisms utilized by cells to respond to specific signals in a precise manner is not well understood. In reconstituted systems there is ample evidence that multiple G proteins can couple to the same sets of receptors and effectors (reviewed in Ref. 1), and in transfected cells a single G␣ subunit can couple to at least three different effector pathways (2). One important mechanism that contributes to regulation of signaling pathways is the existence of proteins that modulate particular points in the pathway. For example, G protein receptor kinases desensitize receptors (such as ␤-adrenergic receptor kinase, (reviewed in Ref. 3) and RGS (regulators of G protein signaling) proteins turn off effector responses by accelerating the GTPase activity of G␣ subunits (reviewed in Ref. 4). Since some G protein family members are predominantly expressed in specific tissues, cell type-specific modulators of G protein signaling are likely to exist.
G␣ o is a member of the pertussis toxin family of G␣ subunits and is predominantly expressed in the central nervous system and heart. Although G␣ o comprises 0.2-0.5% of brain particulate protein (5), many of its functions are yet to be defined. G␣ o couples to several well characterized receptors in the brain and can regulate both N-type Ca ϩ2 channels as well as some K ϩ channels (see Ref. 6 and references therein). In addition, G␣ o can be regulated by neuromodulin (GAP43 (growth cone-associated protein)) in developing neurites (7). Knockout of neuromodulin in mice causes significant abnormalities in neuronal pathfinding (8), but G␣ o knockout mice have anatomically normal brains. Despite the normal central nervous system anatomy in the G␣ o knockout mice, they develop a spectrum of neurologic abnormalities (tremor and impairments of motor control and behavior) and shortened survival (6,9). We utilized the yeast two-hybrid interaction system to search for unique modulators of G␣ o that are expressed in the central nervous system. An interaction between G␣ o and Pcp2 (Purkinje cell protein-2), a protein of unknown function expressed in Purkinje cells and retinal bipolar neurons was identified. Although G␣ o and Pcp2 have not yet been definitively colocalized in cerebellar Purkinje cells, this interaction was confirmed in vitro and by coimmunoprecipitation from transfected cells. Furthermore, Pcp2 can function as a nucleotide exchange factor by stimulating GDP release from G␣ o .

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Screening-G␣ o cDNA in pBS (previously described in Ref. 10) was cloned into EcoRI/SalI sites of PAS2-1 (CLON-TECH), a vector that encodes GAL4 DNA-binding domain. G␣ o -PAS2-1 was used as a "bait" to screen a mouse brain library in pACT (CLON-TECH). Both G␣ o -PAS2-1 and the mouse brain cDNA libraries were cotransformed into Y190, a yeast lacZ/HIS3 reporter strain, using standard methods. The transformed mix was screened for growth on plates containing selective medium (synthetic complete medium lacking tryptophan, leucine, and histidine in the presence of 25 mM 3-aminotriazole) and incubated at 30°C for 8 -12 days. His ϩ colonies were screened for ␤-galactosidase activity using a filter lift assay, and positive "blue" colonies (␤-galactosidase-positive) were further confirmed by a yeast mating assay. Individual plasmids were transformed into Escherichia coli by electroporation and plasmids analyzed by restriction analysis and dideoxynucleotide sequencing. GenBank TM data bases * This work was supported by the National Institutes of Health and by a pilot award from the Harvard Digestive Disease Center (to B. M. D.). 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 1 The abbreviations used are: G proteins, guanine nucleotide-binding proteins; GTP␥S, guanosine 5Ј-(␥-thio)triphosphate; PAGE, polyacrylamide gel electrophoresis; GST, glutathione S-transferase; PBS, phosphate-buffered saline; HA, hemagglutinin.
were screened with each sequence using BLAST analysis.
In Vitro Binding Assay-Pcp2 was excised from pACT2 using EcoRI/ XhoI and cloned into pGEX4 -2 (Amersham Pharmacia Biotech). The resulting Pcp2-pgex cDNA and pGEX4 -2 without insert (GST alone) were transformed into E. coli, and the expressed proteins were purified from bacterial pellets after induction with isopropyl-1-thio-␤-D-galactopyranoside. Bacterial lysates were incubated with glutathione-agarose beads (2 ml) and rotated at 4°C for 30 min. After centrifugation, GST-Pcp2 and GST alone were eluted from beads using glutathione elution buffer for 10 min at room temperature (Amersham Pharmacia Biotech). Samples were recentrifuged, and the resulting supernatant containing the fusion proteins was analyzed according to the Bradford method to determine protein concentration. G␣ o (subtype 1) and G␣ i2 in pBS (G␣ i2 plasmid construction described in Ref. 11) and G␣ s in pcDNAI (from ATCC, Manassas, VA) were used for in vitro translation. Labeled G␣ subunits were made using 1 g of cDNA, appropriate RNA polymerase in a coupled rabbit reticulocyte translation system (TNT system, Promega, Madison WI) plus [ 35 S]methionine (NEN Life Science Products; 20 Ci/reaction) as described previously (11). [ 35 S]Methionine-labeled G␣ subunits were analyzed by SDS-PAGE and autoradiography, and the amounts were normalized by densitometric analysis of translated products (NIH Image 1.61/fat, Wayne Rasband, NIH, Bethesda, MD). Equivalent amounts of fusion proteins (GST or GST-Pcp2) were incubated with glutathione-agarose beads for 30 min at room temperature, followed by incubation with equivalent amounts of in vitro translated [ 35 S]methionine-labeled G␣ o , G␣ s , or G␣ i2 in PBS with 0.05% Triton X-100 overnight at 4°C with rocking. Beads were centrifuged, washed with PBS with 0.05% Triton X-100 three times, eluted with SDS sample buffer, and analyzed by SDS-PAGE and autoradiography.
Coimmunoprecipitation from Transiently Transfected COS Cells-To preserve the HA epitope located at the N terminus of Pcp2 in pACT2, the plasmid was cut with BglII and filled with Klenow to generate a blunt end. Following XhoI digest, the fragment was cloned into the EcoRV/XhoI sites of pcDNA3 (Invitrogen). G␣ o in pcDNA3 (12) was transfected into COS-7 cells alone or in combination with Pcp2 using LipofectAMINE TM (Life Technologies, Inc.) according to the manufacturer's protocol. At 72 h after the transfection, cells were washed with ice-cold PBS and lysed for 30 min in lysis buffer (50 mM Hepes, pH 7.5, 6 mM MgCl 2 , 1 mM EDTA, 75 mM sucrose, 2.5 mM benzamidine, 1 mM dithiothreitol, and 1% Triton X-100). Lysates were cleared by low speed centrifugation, and the supernatant was incubated with the HA-specific monoclonal antibody 12CA5 (1:100) overnight at 4°C. Protein A-Sepharose (Sigma) was added for 1 h, and the samples were rocked at 4°C. Samples were then centrifuged, and the pellets were washed three times with PBS with 0.05% Triton X-100. The precipitated proteins were eluted with SDS sample buffer and analyzed by SDS-PAGE and Western blotting using a rabbit polyclonal anti-G␣ o antibody (5), and bands were visualized by chemiluminescence (Pierce).
Determination of k cat and k off -Bovine brain G␣ o was kindly provided by E. Neer (Harvard Medical School) and used for kinetic analysis in the presence and absence of Pcp2. Pcp2 was prepared from GST fusion proteins by cleavage with thrombin at 2.5 units/mg protein for 1 h at room temperature and separated from GST by incubation with glutathione-agarose beads. Samples were concentrated and protein concentration determined (Bradford). Single turnover GTP hydrolysis (k cat ) was determined by incubating G␣ o (50 nM) with 1 M [␥-32 P]GTP (5500 cpm/pmol) for 20 min in the presence (1:1 molar ratio) or absence of Pcp2 in buffer A (50 mM Tris, pH 7.6, 5 mM EDTA, 1 mM dithiothreitol, and 0.1% Triton X-100). The hydrolysis reaction was started by addition of MgCl 2 (final concentration, 10 mM) and 100 M GTP. Aliquots were diluted into 1 ml of 5% (w/v) trichloracetic acid in 5% charcoal and counted as described previously (13). The amount of ␥-32 P released at each time point was fit to an exponential function using GraphPad Prism. For determining k off , G␣ o (2 pmol) in the presence and absence of Pcp2 or GST (1:1 ratio with G␣ o ) was incubated in buffer A with 10 mM MgCl 2 , 1 M GTP [␣-32 P]GTP (specific activity 1.1 ϫ 10 5 cpm/pmol), and 10 g/ml bovine serum albumin for 20 min at room temperature. The stoichiometry of binding was approximately 60%, and the amount bound prior to initiating nucleotide exchange was set at 100%. GDP dissociation from G␣ o was initiated by the addition of 1 mM GDP, and aliquots were filtered onto nitrocellulose, washed, and counted. The data were fit to a one phase exponential decay using GraphPad Prism (San Diego, CA).

RESULTS AND DISCUSSION
The yeast two-hybrid interaction system detects low affinity protein interactions and has been successfully used for finding novel proteins that interact with G protein ␣ subunits (14). The initial screen yielded 34 ␤-galactosidase-positive clones, and this number was reduced to six positive clones after yeast mating controls. These six clones were sequenced and searched in GenBank TM (BLAST) data bases. Three of these cDNAs had no sequence homology to known genes, and one was identified as lactate dehydrogenase. One of the remaining genes was a previously identified human mosaic protein, LGN, that was discovered in a yeast two-hybrid screen using G␣ i2 as bait (15). The other remaining gene was a full-length clone of PCD5 (now called Pcp2) (16) that had been initially identified as a candidate gene important for Purkinje cell development. In pcd mice, Purkinje cells develop normally but then begin to degenerate at 15-18 days after birth leading to the development of ataxia (17).
To determine whether G␣ o could interact with Pcp2 in an independent assay, GST pull-down experiments were performed. G␣ subunits were translated and [ 35 S]methionine-labeled in vitro and then incubated with equivalent amounts of GST or GST-Pcp2 fusion proteins. Fig. 1 shows that in vitro translated G␣ o binds to GST-Pcp2 (lane 2), whereas no significant interaction was seen with the GST control protein (black arrow, lane 1). We consistently detected approximately 10% of in vitro translated G␣ o coprecipitating with GST-Pcp2. Nonspecific interactions of G␣ o with GST were less than 1%. We next asked whether the conformation of G␣ o affected the interaction with GST-Pcp2. G␣ o was preincubated with GTP␥S (nonhydrolyzable GTP analogue) or GDP prior to binding. As shown in Fig. 1, the amount of G␣ o that associates with Pcp2 is similar irrespective of the nucleotide bound to G␣ o . To address the issue of which G␣ subunits could interact with Pcp2, two other G␣ subunits expressed in the central nervous system were studied. G␣ i2 (pertussis toxin family member; ϳ70% amino acid identity to G␣ o ) and G␣ s (cholera toxin family member; ϳ40% amino acid identity to G␣ o ) were characterized in pulldown experiments with GST-Pcp2. Equal amounts of 35 S-labeled G␣ i2 and G␣ s were translated in vitro and used for binding to GST proteins, and the amount of nonspecific binding between G␣ i2 or G␣ s with GST was similar to that seen with G␣ o (not shown). Fig. 1 shows that in vitro translated G␣ i2 interacts with GST-Pcp2, but G␣ s is barely detectable in GST-Pcp2 precipitates (lane 6, 52 kDa, open arrow). Taken together, these results suggest that the related pertussis toxin family To look for evidence that G␣ o and Pcp2 could interact in cells, cotransfection studies were performed in COS cells. COS cells were transfected with G␣ o , Pcp2, empty expression vector pcDNA3 (PC) alone or in combination. Cell lysates were immunoprecipitated using the 12CA5 antibody directed toward the hemagglutinin epitope on the N terminus of Pcp2 and then analyzed by Western with anti-G␣ o antibody (Fig. 2). COS cells do not normally express G␣ o , and in cells transfected with vector (PC), Pcp2, or G␣ o and then immunoprecipitated with 12CA5 there is no detectable G␣ o immunoreactivity (Fig. 2,  arrow). However, when both G␣ o and Pcp2 are cotransfected into the same cells, a fraction of G␣ o is found in association with Pcp2 (Fig. 2, lane 4). We were unable to detect Pcp2 in immunoprecipitates of G␣ o , presumably due to disruption of the association with G␣ o by the stringent detergent conditions necessary to immunoprecipitate G␣ o (18). Attempts to detect G␣ i2 in Pcp2 immunoprecipitates were unsuccessful, although the level of G␣ i2 expression was much lower than for G␣ o . The observation that G␣ o and Pcp2 interact in an intact cell raises the possibility that Pcp2 could be a modulator of G␣ o .
There are several possible mechanisms for regulation of G␣ subunits by accessory proteins, including effects on nucleotide binding or GTP hydrolysis. To address the possibility that Pcp2 modulates the enzymatic properties of G␣ o , we measured k cat and k off of bovine brain purified G␣ o (5) in the presence and absence of Pcp2. Fig. 3 shows a representative experiment, and the results are summarized in Table I. The k cat and k off values obtained for G␣ o are similar to literature values (19), and there was no significant difference in k cat of G␣ o in presence of Pcp2 ( Fig. 3a and Table I). However, as seen in Fig. 3b and Table I, there is significant stimulation of k off in the presence of Pcp2. In experiments simultaneously comparing k off of G␣ o in the presence and absence of Pcp2 or GST, there was 5.2 Ϯ 0.5-fold stimulation of k off in the presence of Pcp2 (n ϭ 7). Incubating G␣ o with GST had no effect on k off (Fig. 3b and Table I), and including bovine serum albumin in the binding buffer also had no effect. The increase in k off of G␣ o in the presence of Pcp2 is similar to the observed increases seen with G␣ subunits reconstituted with receptors plus agonists (4 -6-fold increases in k off ) (20). This finding suggests that Pcp2 could mimic a receptor by stimulating GDP release. A family of proteins that promote nucleotide exchange have been described for monomeric G proteins such as Ras (21), but only neuromodulin has been shown to affect nucleotide binding to G␣ o . This mechanism of activation is likely to be distinct from Pcp2 because neuromodulin stimulates GTP␥S binding through its N-terminal domain, which is homologous to the cytoplasmic tail of G protein-coupled receptors (7).
The gene for Pcp2 is located on mouse chromosome 8 and encodes a cytosolic 99-amino acid protein without significant homology to other proteins (including G protein-coupled recep- . The fraction of GTP hydrolyzed was measured for bovine brain purified G␣ o in the presence of [␥-32 P]GTP, and k cat was determined by fitting the data to a single exponential association function (GraphPad Prism). The results are summarized in Table I. b, the rate constant for GDP release (k off ) from G␣ o was determined alone (E) or in the presence of PCP2 (q) or an equivalent amount of GST (ϫ). The percentage of GDP remaining bound to G␣ o was determined by equilibrating 1 M [␣-32 P]GTP for 20 min and initiating nucleotide exchange with 1 mM GDP. The fraction of GDP bound to G␣ o was measured over time. The data were fit to a single exponential dissociation, and the results are summarized in Table I. The fold increase in k off for this experiment was 5.9 (the mean of all experiments (n ϭ 7) was 5.2 Ϯ 0.5). tors). There is some amino acid sequence similarity to the c-sis/PDGF2 gene, but the implications of this are unknown (22). The expression profile of Pcp2 is consistent with a role in Purkinje cell development (23), but in mice without Pcp2 expression, Purkinje cells develop normally (presumably through other compensatory mechanisms) (24,25). The function of Pcp2 and the mechanism(s) for increased Purkinje cell apoptosis in pcd mice remains unknown. Several genes in pcd mice have altered expression patterns including up-regulation of the early response genes fos and jun and down-regulation of the anti-apoptosis gene bcl-2 (26). Several G protein ␣ subunits trigger apoptosis (27), and G␣ o may be involved in apoptosis triggered by a mutated amyloid precursor protein in Alzheimer's disease (28). The results described in these studies suggest that Pcp2 may be involved in a signal transduction pathway involving G␣ o , but it remains to be determined whether this pathway relates to apoptosis. In addition, our results do not exclude important interactions of Pcp2 with other G␣ subunits, particularly G␣ i1 , which is highly expressed in the brain. Increasingly, novel functions (and locations) for G proteins are being identified (such as in the Golgi, in the endoplasmic reticulum, and on intracellular vesicles) where they regulate protein processing and vesicular targeting (29 -31). Because G protein-coupled transmembrane receptors have not been identified in many of these locations, G␣ subunits may be activated by nucleotide exchange factors. Pcp2 may be such a factor for G␣ o in cerebellar Purkinje cells, and future studies will define the functional consequences of this interaction.