The α-subunit of the trimeric GTPase Go2 regulates axonal growth


  • Jens Baron,

    1. Center for Anatomy, Institute for Integrative Neuroanatomy, Functional Cell Biology, Charité-Universitätsmedizin Berlin, Germany
    Search for more papers by this author
    • Both authors contributed equally to the article.
  • Christian Blex,

    1. Center for Anatomy, Institute for Integrative Neuroanatomy, Functional Cell Biology, Charité-Universitätsmedizin Berlin, Germany
    Search for more papers by this author
    • Both authors contributed equally to the article.
  • Astrid Rohrbeck,

    1. Institute of Toxicology, Hannover Medical School (MHH), Germany
    Search for more papers by this author
  • Sivarama Krishna Rachakonda,

    1. Center for Anatomy, Institute for Integrative Neuroanatomy, Functional Cell Biology, Charité-Universitätsmedizin Berlin, Germany
    Search for more papers by this author
  • Lutz Birnbaumer,

    1. Laboratory of Neurobiology, Division of Intramural Research, National Institute of Environmental Health Sciences, Research Triangle Park, North Carolina, USA
    Search for more papers by this author
  • Gudrun Ahnert-Hilger,

    Corresponding author
    • Center for Anatomy, Institute for Integrative Neuroanatomy, Functional Cell Biology, Charité-Universitätsmedizin Berlin, Germany
    Search for more papers by this author
  • Irene Brunk

    Corresponding author
    • Center for Anatomy, Institute for Integrative Neuroanatomy, Functional Cell Biology, Charité-Universitätsmedizin Berlin, Germany
    Search for more papers by this author

Address correspondence and reprint requests to Irene Brunk or Gudrun Ahnert-Hilger, Institute for Integrative Neuroanatomy, Philippstr. 12, 10115 Berlin, Germany.



The Goα splice variants Go1α and Go2α are subunits of the most abundant G-proteins in brain, Go1 and Go2. Only a few interacting partners binding to Go1α have been described so far and splice variant-specific differences are not known. Using a yeast two-hybrid screen with constitutively active Go2α as bait, we identified Rap1GTPase activating protein (Rap1GAP) and Girdin as interacting partners of Go2α, which was confirmed by co-immunoprecipitation. Comparison of subcellular fractions from brains of wild type and Go2α−/− mice revealed no differences in the overall expression level of Girdin or Rap1GAP. However, we found higher amounts of active Rap1-GTP in brains of Go2α deficient mutants, indicating that Go2α may increase Rap1GAP activity, thereby effecting the Rap1 activation/deactivation cycle. Rap1 has been shown to be involved in neurite outgrowth and given a Rap1GAP-Go2α interaction, we found that the loss of Go2α affected axonal outgrowth. Axons of cultured cortical and hippocampal neurons prepared from embryonic Go2α−/− mice grew longer and developed more branches than those from wild-type mice. Taken together, we provide evidence that Go2α regulates axonal outgrowth and branching.

Abbreviations used

GAL4 activating domain


activator of G-protein signaling


GAL4 DNA binding domain


GTPase-activating protein


guanine nucleotide exchange factor


Gα-interacting vesicle-associated protein


G-protein signaling modulator


G-Protein regulated inducer of neurite outgrowth




lysed pellet 1, synaptic vesicle


lysed pellet 2, synaptic vesicle


microtubule associated protein


neurofilament protein


optical density




regulator of G-protein signaling






vascular endothelial growth factor

Heterotrimeric G-proteins, consisting of an α subunit and a βγ dimer, couple cell surface receptors to intracellular second messenger systems (Birnbaumer 2007). In addition, they are also located on secretory granules and synaptic vesicles (Ahnert-Hilger et al. 1994; Pahner et al. 2003; Takamori et al. 2006). There are four families of Gα-subunits, Gsα, Gi/oα, Gq/11α, and G12/13α (Simon et al. 1991). Goα belongs to the Gi/o subfamily and represents together with Giα up to 1.5% of membrane protein in mammalian brains (Neer et al. 1984; Sternweis and Robishaw 1984). Goα comprises two splice variants Go1α and Go2α. Go1α has been shown to be selectively involved in the light response of ON-bipolar cells (Dhingra et al. 2002). Vesicle-associated Go2α inhibits vesicular monoamine transporter 1 and 2 (Ahnert-Hilger et al. 1998; Holtje et al. 2000) and affects the chloride dependence of vesicular glutamate transporters (Winter et al. 2005). Furthermore, Go2α has been shown to inhibit adenylyl cyclase activity in reconstituted assay systems (Kobayashi et al. 1990), to reduce voltage-sensitive calcium currents in neurons of the pond snail helisoma (Man-Son-Hing et al. 1992), and to regulate glucose-induced release of insulin (Wang et al. 2010). Besides these Go2α-specific effects, most of the reported Goα interactions refer only to Go1α. Accordingly, Goα/Go1α has been shown to interact with regulator of G-protein signaling (RGS)14, RGS17 and G-Protein-Regulated-Inducer-of Neurite-Outgrowth 1 and 2 (GRIN 1 and 2) (Chen et al. 1999; Traver et al. 2000; Nakata and Kozasa 2005), Rap1GAP, IP6 and Purkinje cell protein 2 (PCP2) (Jordan et al. 1999; Luo and Denker 1999), whereas proteins directly interacting with Go2α have not been identified at this time.

There is growing evidence for a cross talk between heterotrimeric and small GTPases. In this respect, it has been shown that the activities of GTP exchanging factors (GEF) and GTPase activating proteins (GAPs) are modulated by heterotrimeric G proteins (Aittaleb et al. 2011). In neuroblastoma cell lines activated Go1α has been shown to modulate neurite outgrowth by interacting with Rap1 via Rap1GAP (He et al. 2006) or Rit (Kim et al. 2008); however, the role of both Goα splice variants was not differentially regarded.

In this study, we performed a yeast two-hybrid screen using constitutively active Go2α (Go2αQ205L) as well as Go1α (Go1αQ205L) and Gqα (GqαQ209L) for comparison. We found that Rap1GAP and the non-receptor GEF Girdin, also known as Gα-interacting vesicle-associated protein (GIV) (Garcia-Marcos et al. 2009), interact with both Go1αQ205L and Go2αQ205L with a preference for the Go2α-subunit. We provide evidence that the lack of Go2α has an impact on levels of Rap1GTP, which is regulated by Rap1GAP, modulating axonal growth.

Materials and methods


Mouse monoclonal antibodies raised against both Goα splice variants (clone 101.1) or specifically recognizing Go2α (clone 101.4) were previously described (Winter et al. 2005). Antibodies against Goα (mouse), Girdin (rabbit), and Rap1 (rabbit) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). The anti-Goα antibody from Santa Cruz has been previously shown to detect only Go1α (Brunk et al. 2008). Antibodies against Rap1GAP (rabbit monoclonal and polyclonal) were obtained from Abcam (Cambridge, UK) and Calbiochem (Merck, Darmstadt, Germany), respectively.

Monoclonal antibodies against synaptophysin (Jahn et al. 1985) and synaptobrevin and a rabbit polyclonal antibody against actin were obtained from Synaptic Systems (Göttingen, Germany) or Sigma-Aldrich (St. Louis, MO, USA), respectively.

Neuronal cultures were stained by a mouse monoclonal antibody against neurofilament protein (NFP; 200 kDa) and a rabbit polyclonal antiserum against microtubule associated protein 2 (MAP2), both from Millipore (Billerica, MA, USA) to identify axons or dendrites, respectively (Ahnert-Hilger et al. 2004).

Secondary antibodies, horse anti-mouse and goat anti-rabbit conjugated to horseradish peroxidase, were purchased from Vector Laboratories (Burlingame, CA, USA). Secondary antibodies conjugated to fluorochromes (goat anti-mouse IgG and goat anti-rabbit IgG, either Cy2 or Cy3) were obtained from Dianova (Hamburg, Germany).


Go1α and Go2α splice variant-specific deletion mutants were bred and genotyped as described (Jiang et al. 1998; Dhingra et al. 2002). For all experiments, except for hippocampal cultures, mutant and wild-type littermates were obtained by interbreeding of heterozygous parents. Institutional approval (approval number T0119/11) was obtained for killing the animals used in this study.

Molecular cloning and yeast two-hybrid experiments

A mouse brain cDNA-library cloned into the vector pACT2 (Clontech-Takara Bio Europe, Saint-Germain-en-Laye, France) was amplified in E.coli BNN132 according to the manufacturer's instructions. Afterward, plasmid-DNA was isolated using the Qiagen Plasmid Giga Kit (Qiagen GmbH, Hilden, Germany).

cDNAs of Go2αQ205L, Go1αQ205L as well as GqαQ209L were cloned into pAGA2 (Hsu et al. 1990; Wei et al. 1991) and then were amplified by PCR using the following primer pairs:







Gα subunit amplicons were cloned into the multiple cloning site of the pGBKT7 plasmid using EcoRI/SalI and EcoRI/XhoI restriction sites, and the absence of PCR-induced mutations was confirmed by DNA sequencing. Plasmid expression in yeast yields to bait fusion proteins in which the Gal4-DNA-binding domain (DNA-BD) is fused to the N-terminus of Gα subunits.

The yeast strain AH109 was sequentially transformed using pGBKT7Go2Q205L and the mouse brain cDNA-library. The transformed cells were grown in quadruple-dropout-selection-medium (QDO) containing 1 mM 3-aminotriazole and lacking tryptophan, leucine, adenine, and histidine at 30°C for up to 2 weeks. Colonies were re-grown three times using the same medium supplemented with 40 μg/mL X-α-Gal (5-Bromo-4-chloro-3-indolyl-α-d-galactopyranoside) to detect α-galactosidase expression. Plasmid-DNAs of α-galactosidase positive colonies were isolated using the Zymoprep II Yeast Plasmid Minipreparation Kit (Zymo Research Corporation, Orange, CA, USA) and amplified after transformation of E. coli DH5α cells. Finally, after E. coli plasmid preparation, library cDNAs cloned into pACT2-vector were sequenced and analyzed using the BLAST-tools of the European Bioinformatics Institute (EBI).

To confirm the protein interaction and to analyze the interaction specificity of the identified clones, yeast mating experiments were performed. To this end, the yeast AH109 strain mating type a was retransformed with the purified plasmid-DNA encoding identified interacting proteins (preys). The yeast Y187 strain mating type α was transformed using either control-vector-DNA or the pGBKT7 vector carrying cDNAs encoding constitutive active Go1α, Go2α, or Gqα cloned into EcoRI/SalI or EcoRI/XhoI restriction sites (baits). Yeast strains were then grown on SD (synthetic dropout) – medium lacking leucine or tryptophan for selection of positive co-transformants. After mating in 200 μL YPD-medium over night at 30°C, diploid cells were selectively grown on SD-Leu-Trp-medium for 3 days. Afterward, growth of yeast cells was analyzed on QDO-medium-plates as described above.

For the p-nitrophenyl-α-d-galactopyranoside (PNP-α-Gal) assay, samples were prepared by yeast mating as described above by crossing cells of the Y187 strain mating type α bearing the bait plasmids pGBKT7-Go2αQ205L, -Go1αQ205L, -GqαQ209L, -Lamin C, -p53, or the empty vector pGBKT7 with cells of the AH109 strain mating type a, bearing the prey plasmids pACT2-Girdin, -Rap1Gap, Synembryn, pGADT7-T, or the empty pGADT7 vector (Table 1). After mating and selection of diploid cells on minimal agar lacking Trp and Leu, 2 mL of the same medium or high stringency SD medium (SD –Trp –Leu –His –Ade +1 mM 3-AT) were inoculated with one large or up to three small colonies expressing the indicated pair of proteins, and incubated overnight at 30°C while shaking at 250 rpm. The cell density of the suspension was adjusted to an OD600 of between 0.4 and 1.0. One milliliter of yeast cells were then collected by centrifugation at 10 000 g for 2 min. Three 16 μL aliquots of each supernatant as well as of medium for blank values were mixed with 48 μL of assay buffer containing 33.3 mM PNP-α-Gal and 0.33 M NaOAc, pH 4.5 in a 96 well microtiterplate and incubated for 1 h at 30°C. A p-nitrophenol standard was run afterward. The reaction was stopped by adding 136 μL of stop solution containing 1 M Na2CO3 and color development was determined at 405 nm using a microtiterplate ELISA reader (Anthos Labtec, Salzburg, WALS, Austria; HT-2). Units of the α-galactosidase were calculated by the following method:

display math
Table 1. Mating pairs of transformed yeast cells
  1. Indicated are the plasmids with which yeast cells of the respective strain were transformed. Transformed Y187 yeast cells (mating type α were mated with transformed AH109 yeast cells (mating type a) as shown.

21pGBKT7-p53pGADT7-large T antigen

OD405, optical density of p-nitrophenol after enzymatic reaction; t, elapsed time of incubation (min); Vf, final volume of assay (200 μL); Vi, volume of culture medium supernatant added (16 μL); OD600, optical density of overnight culture; DF, dilution factor of OD600; ε × b, p-nitrophenol molar absorption at 405 nm multiplied by the light path (cm) and the proportionality constant of a linear standard regression line.

Significance was tested from five (Girdin interaction) or four (Rap1GAP interaction) independent experiments using the Kruskal–Wallis test on the basis of a significance level of < 0.05.

Immunoblot analysis

Subcellular fractions were prepared from mouse brains as described previously (Huttner et al. 1983; Becher et al. 1999). Samples were analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and western blot techniques using the ECL detection system (GE Healthcare, Munich, Germany). For quantification, ECL-processed films were scanned and protein bands were densitometrically measured using the Labimage 1D program (KAPELAN, Halle, Germany) ensuring that signals were in the linear range of the ECL detection system. Comparative quantification of protein samples from wild-type mice and deletion mutants was performed in the same gel using actin or synaptophysin as internal controls.


Synaptic vesicle preparations (LP2; 1 mg) were extracted with 1 mL extraction buffer containing 140 mM KCl, 2 mM EDTA, 20 mM HEPES-KOH (pH 7.4), and 1% (v/v) Triton X-100; and then incubated for 1 h at 4°C under rotation, followed by centrifugation at 1300 g for 5 min. Two hundred microliter of each extract were then immunoprecipitated using the following antibodies: rabbit anti-Rap1GAP, rabbit anti-Girdin, mouse anti-Go2α, normal mouse IgG, normal rabbit IgG. All antibodies were diluted to give a final immunoglobulin concentration of 2–4 μg per sample depending on the antibody charge. Immunoprecipitation was performed using Dynabeads™ (Invitrogen, Darmstadt, Germany). Following the binding of antibodies to 50 μL beads (incubation for 1 h at 20°C while rotating) and washing, 200 μL of brain extract was added and samples were incubated over night at 4°C. The supernatants were kept for further analysis, whereas the beads were washed three times in extraction buffer. Finally, the bead pellets were resuspended in loading buffer and boiled for 10 min. Supernatants and pellets were analyzed by SDS-PAGE and western blotting.

Rap1 pull-down assay

The Rap1-binding domain of RalGDS (RalGDS-RBD) was expressed as a GST fusion protein in E. coli and purified by affinity chromatography using glutathione-sepharose. Frozen mouse brains (males) were homogenized on ice in a Potter-Elvehjem homogenizer, in freshly prepared ice-cold buffer (150 mM NaCl, 50 mM Tris pH 7.2, 5 mM MgCl2, 1 mM phenylmethanesulfonylfluoride, 5 mM dithiothreitol, 1% NP-40 and protease inhibitor cocktail), by hand using 8 to 10 up and down strokes of the Teflon pestle and then sonicated. The solubilized tissue was spun at 13 000 g for 10 min at 4°C, DNAse digestion was then performed at 20°C for 15 min and lysates were clarified by centrifugation at maximal speed for 10 min at 4°C. After 15 min, the GTP-bound state of Rap1 was stabilized with 60 mM MgCl2. Glutathione S-transferase (GST)-tagged Ral-GDS beads were added to the supernatants and incubated for 1 h (4°C). The beads were washed three times, and bound proteins were solubilized by incubation with loading buffer at 95°C for 10 min. Samples were subjected to SDS-PAGE and western blot analysis using a polyclonal antibody raised against Rap1 (Cell Signalling, Danvers, MA, USA). Finally, all signals were analyzed densitometrically using the KODAK 1D software Image Analysis Software (Eastman Kodak Company, Stuttgart, Germany) and normalized to β-actin signals. Pull-down results were statistically tested with five animals per genotype using a two-sample paired t-test on the basis of a significance level of p < 0.05.

Cell culture

Neurons were prepared from fetal brains of wild type, heterozygous, and Go2α−/− mice or from wild type and Go1α−/− mice at embryonic day 16 (E16) as described previously (Ahnert-Hilger et al. 2004). Pieces of the hippocampus or neocortex were rinsed with phosphate-buffered saline (PBS), then with dissociation medium [modified Eagle medium (MEM) supplemented with 10% fetal calf serum, 100 IU/L insulin, 0.5 mM glutamine, 100 U/mL penicillin/streptomycin, 44 mM glucose, and 10 mM HEPES buffer] and then dissociated mechanically. Following centrifugation, cells were resuspended in starter medium (serum-free neurobasal medium supplemented with B27, 0.5 mM glutamine, 100 U/mL penicillin/streptomycin, and 25 μM glutamate) and plated at a density of 2 × 104 cells/well on poly-l-lysine/collagen pre-coated glass coverslips. All ingredients were obtained from Gibco/BRL Life Technologies (Eggenstein, Germany).

For hippocampal cultures, wild type and Go2α−/− strains were bred separately and hippocampi of each genotype were collected. For littermate neocortical cultures, wild type (129/Sv x C57BL/6) and Go2α splice variant-specific mutants or wild type and Go1α splice variant-specific mutants were cross-bred, and the resulting heterozygous offsprings were used to generate littermates with the same genetic background. Neocortical neurons of each individual embryo were cultured separately and the genotype of the individual cultures was determined.

Immunocytochemistry and morphometric analysis of neuronal cultures

After 5 days in vitro (DIV) neurons were fixed with 4% formalin for 15 min and subsequently permeabilized for 30 min at 20°C using 0.3% Triton X-100 dissolved in PBS. For confocal laser scanning microscopy, neurons were stained with the indicated primary antibodies overnight at 4°C. For morphometric analysis, antibodies against NFP200 and MAP2 were used to mark axons and dendrites, respectively. After washing in PBS, secondary antibodies were applied for 1 h at 20°C. After three rinses with PBS and one rinse with water, coverslips were mounted on glass slides for microscopic analysis of fluorescent labeling.

For confocal laser scanning microscopy we used a Leica DMRE microscope (Leica Microsystems GmbH, Heidelberg, Germany), with an ×63 oil immersion objective lens, with excitation frequencies of 488 nm (argon laser) and 543 nm (helium-neon laser). Image acquisition and analysis was performed using the built-in Leica Confocal Software.

Image acquisition preceding morphometric analysis was performed by a Leica DMLB microscope using an ×40 objective lens. Images were captured at a resolution of 1024 × 1024 pixels. Total length and overall number of branching nodes of axons and dendrites were analyzed morphometrically using the Neurolucida software (MicroBrightField, Williston, VT, USA). The parameter ‘axon length’ represents the integral total length of all visible parts of an axon, including its higher order branches (Ahnert-Hilger et al. 2004). Experiments were carried out twice for hippocampal and twice for neocortical neurons from cultures prepared on the same day. Typically, five coverslips were prepared per genotype and six neurons were evaluated on each coverslip. Data from five coverslips (30 individual neurons) were pooled and given as means ± SD. The data displayed in the graphs of Fig. 5 refer to a single, representative experiment, if not mentioned otherwise. Genotype related differences were statistically analyzed using the Kruskal–Wallis and Mann–Whitney-U-test, on the basis of a significance level of p < 0.05.


Interaction partners of constitutively active Go2α

To identify interaction partners of Go2α, a yeast two-hybrid screen was performed using a mouse brain cDNA-library as prey and constitutively active Go2αQ205L as bait. Screening of 3.5 × 106 clones initially yielded 20 different interacting proteins, which were reduced to 11 after performing yeast mating controls (Table 2). Proteins already identified as important regulators of Gα- signaling were found to interact with Go2α. These include regulators of G-protein signaling (RGS-proteins) 14, 17, 19, and 20, as well as G-protein signaling modulator 1 (Gpsm1 also called AGS3), and Gpsm2 also called LGN, which have several GoLoco motifs, and potentially positively or negatively regulate G-protein signaling (Kerov et al. 2005). Known binding partners of Go1α, like Rap1GAP (Jordan et al. 1999), or synembryn (Tall et al. 2003), were also identified as was Girdin, which has been described to bind to Goα, Gsα, and Gi1-3α before, however, the Goα splice variant involved was not elucidated (Le-Niculescu et al. 2005). Moreover, we identified proteins for which no interaction with α-subunits of heterotrimeric G-proteins has been reported so far, such as the Armadillo repeat containing x-linked 1 (ALEX1) (Kurochkin et al. 2001), which may function as a tumor suppressor.

Table 2. Interaction partners of Go2αQ205L
  1. Indicated are interaction partners of the constitutively active mutant of Go2α identified by a yeast two-hybrid screen. Interaction with Go1α and Gqα was investigated to determine selectivity of interaction.

  2. +, interaction; −, no interaction.

Armadillo repeat containing x-linked 1++
G-protein signaling modulator 1++
G-protein signaling modulator 2++
Rap1Gap (KIAA0474)++

Specificity of these interactions was controlled by yeast mating experiments using the AH109 and Y187 strains, carrying the identified interacting partners (prey) and the constitutively active Gα-subunits (baits) fused to GAL4 domains, respectively. To determine isoform specificity of the identified interactions, Go1αQ205L was chosen for a direct comparison with Go2αQ205L. In addition, an interaction with GqαQ209L belonging to the Gq/11-family of Gα-subunits, was analyzed to determine G-protein family specificity (Table 2). We found, that diploid yeast colonies, expressing Rap1GAP or Girdin in combination with Go2αQ205L, grew faster compared with diploid yeast colonies bearing Rap1GAP or Girdin in combination with Go1αQ205L. For Girdin, two partial cDNAs were recovered in the screen (coding for amino acids G1318-S1873 and L1351-S1873), whereas for Rap1GAP three full length clones were recovered. Synembryn also interacted with both Goα splice variants, however, also with Gqα.

To further analyze this effect, we used a quantitative α-galactosidase assay (Fig. 1). Unspecific binding of the Gal4-DNA-activating domain to bait fusion proteins and unspecific interactions of the Gal4-DNA-binding domain to prey fusion proteins were excluded (Fig. 1a). In addition, the absence of α-Gal activation during mating, when using the LaminC-binding domain fusion protein in combination with the respective prey-activating domain fusion protein, served as negative control. As can be seen in Fig. 1b, the Go2αQ205L/Girdin interaction was 2.5-fold larger than the Girdin/Go1αQ205L interaction. On the same lines, the interaction between Rap1GAP and Go2αQ205L (Fig. 1c) was 1.9-fold stronger than with Go1αQ205L. There was no interaction of either protein with the activated form of Gqα (Fig. 1).

Figure 1.

Relative strength of Go2α/Girdin and Go2α/Rap1Gap interaction determined by quantitative α-galactosidase assay. Mating-derived diploid cells were probed for α-Galactosidase (α-Gal) expression and secretion that resulted from interacting Gal4 fusion proteins as indicated. Lamin C acted as a negative control. BD, binding domain; AD, activating domain. (a) The displayed representative experiment illustrates the considerably stronger interaction of Girdin and Rap1Gap with Go2αQ205L compared with Go1aQ205L and the lack of an interaction between GqαQ209L with both preys. In addition, the following negative controls are depicted: Empty prey plasmids were combined with each bait: Go2QL-BD+AD, Go1QL-BD+AD, GqQL-BD+AD. Empty bait plasmids were combined with each prey: BD+Girdin-AD, BD+Rap1GAP-AD. The LaminC containing bait plasmid was combined with each prey, because LaminC is known to interact with almost no other protein: LaminC-BD+AD, LaminC-BD+Girdin-AD, LaminC-BD+Rap1GAP-AD. Empty prey and bait plasmids were combined: BD+AD. (b) Summary of five independent experiments: The Go2αQ205L/Girdin interaction exceeds Girdin´s interaction with Go1αQ205L 2.5-fold; significance was confirmed by the Kruskal–Wallis test (p < 0.02); star denotes significance. There is no interaction of Girdin with GqαQ209L since the level of α-Gal activity is comparable to the negative controls. (c) Summary of four independent experiments: The interaction between Rap1GAP and Go2αQ205L is 1.9-fold stronger than with Go1αQ205L; significance was confirmed by the Kruskal–Wallis test (p < 0.03); star denotes significance. There is no interaction of Rap1Gap with GqαQ209L since the level of α-Gal activity is comparable to the negative controls.

The interaction of Go1α and Go2α with Rap1GAP and Girdin was confirmed by immunoprecipitation (Fig. 2). In agreement with the results obtained from the yeast two-hybrid screen, immunoprecipitation performed with anti-Go2α (Fig. 2a) led to co-precipitation of Rap1GAP and Girdin, whereas precipitation with normal mouse IgG did not yield any specific precipitate. Immunoprecipitation using anti-Girdin (Fig. 2b) and anti-Rap1GAP (Fig. 2c) led to co-precipitation of Go1α and Go2α.

Figure 2.

Co-immunoprecipitation of Girdin and Rap1GAP with Go2α. SN, supernatant; IP, immunoprecipitate. (a) Immunoprecipitates from mouse LP1 derived from total brain using an antibody against Go2α and normal mouse IgG as negative control. Confirming the results from the yeast two-hybrid screen both Girdin and Rap1GAP were co-precipitated with Go2α. Immunoprecipitation using mouse IgG yielded no specific precipitate. (b) Immunoprecipitates from mouse LP2 derived from total brain using a Girdin antibody contained besides Girdin, Go1α and Go2α, confirming the results from the yeast two-hybrid screen. Synaptophysin was not co-precipitated and served as a negative control. (c) In agreement with the results from the yeast two-hybrid screen, the Rap1GAP antibody precipitated Rap1GAP together with Go2α and Go1α using a Triton X-100 extract from mouse synaptic vesicles. Synaptophysin was not co-precipitated and served as a negative control.

Expression of Girdin and Rap1GAP in wild type and Go2α−/− mice

As expected, synaptosomes (P2) from Go2α knock out brains are devoid of Go2α as identified by the subtype-specific antibody (Fig. 3a). Following subcellular fractionation of whole adult brains, Girdin was predominantly found in the LP2 fraction (Fig. 3b). Girdin binds to actin filaments and is found on ER-Golgi transport vesicles (Le-Niculescu et al. 2005), as well as associating with the plasma membrane, depending on its phosphorylation state (Enomoto et al. 2005). Comparison of wild type and Go2α−/− mice did not reveal significant differences in expression levels of Girdin in different fractions derived from whole brains (Fig. 3b).

Figure 3.

Expression of Girdin and Rap1GAP in brains of wild type and Go2α−/− mice. (a) P2 fractions from wild type and Go2α−/− brains were analyzed using antibodies against Go2α (clone 101.4) or against both Goα splice variants (clone 101.1) affirming absence of Go2α in deletion mutants. Synaptobrevin served as an internal control. (b) Subcellular fractions prepared from wild type and Go2α−/− brains were analyzed by western blotting using a Girdin antibody. Expression of Girdin did not differ between the genotypes. The graphs show quantifications of lanes loaded with 10 µg protein. (c) Subcellular fractions prepared from wild type and Go2α−/− brains were analyzed by western blotting using a Rap1GAP antibody. The amount of Rap1GAP was lower in LP1 from Go2α−/− mice than in wild-type LP1. Significance was affirmed by the Kruskal–Wallis-test (p < 0.04); star denotes significance. Regarding the other brain fractions, there were no differences between the genotypes. The graphs show quantifications of lanes loaded with 10 µg protein. (b–c) Quantification is given as relative optical density (OD) and was assessed using synaptophysin as an internal control. Values represent the mean of four animals per genotype ± SD. H, homogenate; P2, synaptosomes; LP1 (lysed pellet 1): LP2, synaptic vesicles.

Similar amounts of Rap1GAP were found in brain homogenates, synaptosomes, and synaptic vesicles of wild type and Go2α−/− mice (Fig. 3c). Only in the LP1 fraction Rap1GAP levels were decreased in comparison to wild-type mice (p < 0.04). The LP1 fraction is known to consist of material from plasma membranes. Go2α may attach Rap1GAP to membrane compartments, so lack of Go2α may cause a shift in sorting. Rap1GAP is known to be enriched in the striatum (McAvoy et al. 2009). Therefore, we analyzed striatal synaptosomes of wild type and Go2α−/− mice; however, no difference of Rap1GAP expression levels was observed (data not shown). Expression patterns of Rap1GAP in the cerebellum, brainstem, striatum, and mesencephalon did not differ between wild type and Go2α−/− mice (data not shown).

Increased amounts of activated Rap1 in Go2α−/− mice

We have shown that constitutively active Go2α interacts with Rap1GAP, the activation of which stimulates the GTPase activity of the Ras-like small G-protein Rap1 to render Rap1 inactive (Rubinfeld et al. 1991). Therefore, we analyzed the expression of Rap1 in different brain fractions and the levels of active GTP-bound Rap1 in brains of wild type and Go2α−/− mice.

Rap1 was comparably expressed in brain fractions of both genotypes (Fig. 4a). Pull-down assays of activated Rap1-GTP from total brains, however, revealed a significantly increased ratio of activated to total Rap1 in Go2α deletion mutants compared with wild-type animals (Fig. 4b). Thus, loss of Go2α results in an increased level of activated Rap1, most likely by decreasing the Rap1GAP activity.

Figure 4.

Amounts of entire and activated Rap1 in brains of wild type and Go2α−/− mice. (a) Subcellular fractions prepared from wild type and Go2α−/− brains were analyzed by western blotting using a Rap1 antibody. The expression of Rap1 did not differ between the genotypes. The graphs show quantifications of lanes loaded with 10 µg protein. Quantification is given as relative optical density (OD) and was assessed using synaptophysin as an internal control. Values represent the mean of four animals per genotype ± SD. H, homogenate; P2, synaptosomes; LP1 (lysed pellet 1): LP2, synaptic vesicles. (b) Increased amount of GTP-bound Rap1 in brains of Go2α−/− mice compared with wild-type littermates as seen in a pull-down assay of activated GTP-bound Rap1. GTP-bound Rap1 was quantified from five male animals per genotype. Values represent the mean (± SEM), star denotes significance (paired t-test, p < 0.02).

Morphometric analysis of neuronal cultures from wild type and Go2α−/− mice

Rap1 is involved in regulation of neurite outgrowth and dendritic development (Jordan et al. 2005; Xie et al. 2005; McAvoy et al. 2009). As the amount of Rap1-GTP is increased in brains of Go2α deletion mutants, compared with wild types, we therefore analyzed axon and dendrite length and branching in neuronal cultures from these genotypes. In Go2α deletion mutants axons grew longer and had more branching nodes in cortical (Fig. 5a left panel, b) and hippocampal (Fig. 5a right panel, c) primary neuronal cultures in comparison to cultures from wild-type mice. Axon lengths in neocortical cultures, from heterozygous mice, were comparable to those from Go2α−/− mice; however, the number of branching nodes was significantly decreased compared with the deletion mutants. Still, there was an increase in branching points in the heterozygous compared with the wild-type neurons (Fig. 5b). These data indicate a gene dose effect of Go2α differentially effecting growth and branching of axons. Dendritic length and branching did not differ between the genotypes (Fig. 5b, c). Taken together, Go2α appears to specifically regulate axonal outgrowth and branching, which is perhaps explained by the interaction of Go2α with Rap1GAP leading to increased Rap1GTP in Go2α deletion mutants suggesting that an enhancement of Rap1 activity might be involved.

Figure 5.

Morphometric analysis of neuronal cultures from wild type and Go2α−/− mice. Total length and overall number of branching nodes from axons and dendrites were analyzed morphometrically in neocortical and hippocampal neuronal cultures from wild type, heterozygous (only for neocortical cultures) and Go2α−/− mice after 5 DIV. The Figure illustrates the results from one representative experiment (see details in material and methods). For each genotype data from 30 neurons were used for calculation of significance. The experiment was performed twice with hippocampal and twice with neocortical neurons. (a) Examples of neocortical (left) and hippocampal (right) neurons from wild type and Go2α−/− mice immunostained for NFP200 (green) and MAP2 (red) to visualize axons and dendrites, respectively. Bar represents 20 µm for upper panels and 10 µm for lower panels. (b) Neocortical neuronal cultures prepared from wild type, Go2α+/− and Go2α−/− littermates were analyzed. Axonal length and branching (upper panels) were significantly increased in Go2α deletion mutants compared with wild-type mice (p < 0.005). Axon length of neurons from heterozygous animals did not differ from axon length in deletion mutants, but from axon length in wild-type animals (p = 0.002). The number of axonal branching nodes per neuron fell between those from wild type and Go2α−/− animals (p < 0.005). Dendrite length and branching (lower panels) did not differ between the genotypes. Significance was tested by the Kruskal–Wallis test; stars denote significance. (c) Hippocampal neuronal cultures prepared from wild type and Go2α−/− mice were analyzed: Axonal length and branching (upper panels) were significantly increased in Go2α deletion mutants compared with wild-type mice (p < 0.005 for both). Dendrite length and branching (lower panels) did not differ between wild type and Go2α−/− mice. Significance was tested by the Kruskal–Wallis test; stars denote significance.

Expression of Girdin, Rap1GAP and Rap1 in embryonic neurons of wild type, Go1α−/− mice and Go2α−/− mice

Girdin, Rap1GAP and Rap1, which interact with Go2α and affect axonal branching, were analyzed in post-nuclear supernatants from embryonic brains of wild type and Go2α−/− mice. Expression levels of Girdin, Rap1GAP and Rap1 were comparable in post-nuclear supernatants of E16 brains from both genotypes (Fig. 6).

Figure 6.

Expression of Rap1GAP, Rap1, and Girdin in embryonic brains from wild type and Go2α−/− mice. (a) Post-nuclear supernatants prepared from embryonic wild type and Go2α−/− brains at E16 were analyzed by western blotting using anti-Girdin, -Rap1GAP, and -Rap1 antibodies. Similar amounts of the proteins were expressed in both genotypes. Detection of actin and synaptophysin served as control. Lanes were loaded with 5 µg, 10 µg, and 15 µg of protein, respectively. The lack of Go2α was analyzed in parallel (b).

Possible co-localization of Rap1GAP (Fig. 7) with Go1α and Go2α was investigated by confocal microscopy of neuronal cultures from wild type and Go1α−/− mice. The latter genotype has been used since the Goα antibody suitable for immunofluorescence detection recognizes both splice variants. The expression pattern of Rap1GAP does not differ in wild type and Go1α−/− neurons (Fig. 7a). Rap1GAP could be detected in the cytosol of perikarya and processes of wild type and Go1α−/− mice. The expression pattern of Rap1GAP also superimposes that of Goα (Fig. 7a, b). Co-localization of Go2α and Rap1GAP is shown by merging immunosignals of anti-Rap1GAP and anti-Goα 1 + 2 in neuronal cell cultures of Go1α−/− mice (Fig. 7a). Determination of fluorescence intensity of both signals clearly confirms their overlay and thereby the co-localization of Go2α and Rap1GAP in processes of primary neurons (Fig. 7b). The same results could be obtained for Rap1 and Girdin (data not shown). Neurons from deletion mutants lacking both Goα splice variants were devoid of immunosignal after incubation with the Goα antibody confirming specificity. Specificity of the Rap1GAP immunolabeling was confirmed by blocking the antibody reaction with the respective antigen used for immunization (Fig. 7c).

Figure 7.

Expression of Rap1GAP in primary neuronal cell cultures from embryonic (E16) brains of wild type and Go1α−/− mice. Expression patterns and co-localization of Rap1GAP and Goα-subunits were investigated by confocal laser scanning microscopy using an antibody against Rap1GAP and an antibody recognizing both Go1α and Go2α. Therefore, co-localization with Go2α can be seen in cultures from Go1α−/− brains. (a) The expression pattern of Rap1GAP does not differ between wild type and Go1α−/− mice. Co-localization of Rap1GAP with Go1α/Go2α in wild-type mice and with Go2α in Go1α−/− mice is demonstrated by the detailed photos on the right. Bars represent 20 µm and 4 µm (right pictures), respectively. (b) Quantification of Rap1GAP- and Goα-immunolabeling in neuronal processes of Go1α deletion mutants confirms the co-localization with Go2α. (c) Specificity of antibody labeling Upper panels: The antibody against Goα (Go2α + Go1α) yields specific immunolabeling in wild-type neurons; no signal can be detected in neurons from Go1α + Go2α−/− mice. Lower panels: Pre-incubation of Rap1GAP antibody with the respective peptide used for immunization blocks immunostaining in neuronal cultures confirming specificity. Bar represents 50 µm.


To our knowledge, this is the first study aiming at the identification of interaction partners specific for the splice variant Go2α which in the brain represents one third of the highly abundant Go heterotrimeric G-protein. Constitutively active Goα variants were chosen as baits for the yeast two-hybrid screen to identify interacting partners and hence are possible effectors of Go2α. We identified several proteins interacting with constitutively active Go2α, which also interacted with constitutively active Go1α. For two of them, Girdin and Rap1GAP, the interaction with Go2α was shown to be stronger compared with Go1α. Physiological relevance of the Rap1GAP-Go2α interaction was confirmed when deletion of Go2α led to an increase of cerebral Rap1GTP and positively affected axonal growth of embryonic neurons.

Interaction of Girdin and Go2αQ205L

Girdin, otherwise known as GIV, is highly expressed in the brain and testis and located on intracellular vesicles, especially Golgi transport vesicles, having been shown to bind to wild-type Gi1-3α, Gsα, Gzα, and Goα subunits (Le-Niculescu et al. 2005). According to our yeast two-hybrid screen, Girdin interacts with both Go2αQ205L and Go1αQ205L, respectively, with a preference for Go2αQ205L. Girdin did not interact with Gqα.

Girdin is believed to be a non-receptor GTP-exchange factor (GEF). In this respect, Girdin may promote activation of vesicular Go2α, which has been shown to regulate vesicular monoamine accumulation (Ahnert-Hilger et al. 1998), although a direct association of Girdin with synaptic vesicles has, so far, not been shown.

Girdin is phosphorylated by Akt, an important regulator of cell survival, growth, and migration; and is associated with actin filaments (Anai et al. 2005; Enomoto et al. 2005). Phosphorylation of Girdin by Akt has been suggested to play an important role in cell motility, progression of malignant neoplasms, and VEGF-mediated angiogenesis (Jiang et al. 2008; Kitamura et al. 2008). In mammalian cells, the role of Girdin in cell migration appears to be regulated by Gi3α dependent on the state of bound GTP. Girdin has been described to be a GEF for Giα (Ghosh et al. 2008; Garcia-Marcos et al. 2009) and to a lesser extent for Go1α (Garcia-Marcos et al. 2010); whether Girdin also functions as a GEF for Go2α is not clear. Given an interaction between activated Go2α and Girdin, Go2α may modulate the GEF function of Girdin for Giα or Go1α. Girdin also interacts with Disrupted-In-Schizophrenia 1 (DISC1) and both play a critical role in post-natal neurogenesis of the dentate gyrus (Enomoto et al. 2009). Thus, Go2α may have an effect on neuronal development via its interaction with Girdin similar to effects reported here regarding axonal growth of embryonic hippocampal and neocortical neurons. Since adult Go2α−/− mice did not show severe morphological abnormalities, the impact of Go2α deletion on the assembly of neuronal network appears to be subtle becoming only evident during examination of specific behavioral tasks. For example, behavioral sensitization was absent in Go2α deletion mutants (Brunk et al. 2008; Brunk et al. 2010).

Interaction of Rap1GAP and Go2αQ205L

Like Girdin, Rap1GAP interacts with both Go2αQ205L and Go1αQ205L. Previous analyses described an interaction of Rap1GAP with wild-type Go1α, but not with the constitutively active Q205L mutant (Jordan et al. 1999) as we have shown. Still, this finding supports the observation from our yeast two-hybrid screen. Furthermore, the yeast two-hybrid α-Gal-assay proves a preferential binding of Rap1GAP to Go2αQ205L over Go1aQ205L. Rap1GAP has also been reported to bind to constitutively active Gzα (Meng et al. 1999).

Rap1GAP is a GTPase activating protein for the small GTPase Rap1 and is most abundant in fetal tissues, but is also found in adult brain (Rubinfeld et al. 1991). There are two human isoforms of Rap1GAP, one of which possesses an N-terminal extension, which reveals a GoLoco domain known to bind Gα subunits. The mouse strain we used, expresses only one isoform containing a GoLoco domain. The expression level of Rap1GAP in different brain fractions was similar in both genotypes except for the LP1 fraction, where Rap1GAP levels were decreased in Go2α−/− compared with wild-type mice possibly because of changes in sorting.

The Rap1GAP effector Rap1 is involved in a variety of cellular functions, including cell proliferation and differentiation via regulation of ERK signaling as well as cell–cell adhesion and migration via integrin mediated pathways (Hattori and Minato 2003). Within the nervous system, regulation of Rap1 activity by Rap1GAP contributes to neurite outgrowth, dendritic development, and dendritic spine plasticity (Chen et al. 2005; Jordan et al. 2005; Xie et al. 2005; McAvoy et al. 2009). Cannabinoid receptor-induced, Rap1-mediated neurite outgrowth has been linked to Go1α (Jordan et al. 2005) with the wild-type Go1α interacting predominantly with human Rap1GAP containing the GoLoco motif. Go1α seems to target Rap1GAP for degradation by the ubiquitin-proteasome system, thereby increasing Rap1 activity (Jordan et al. 2005). Another way by which Rap1GAP and as a consequence Rap1 can be regulated is phosphorylation (McAvoy et al. 2009) leading to inhibition of GAP activity and, thus, to an increased Rap1 activity. In brains of Go2α−/− mice the ratio of active Rap1-GTP to total Rap1 was increased when compared with wild-type littermates. Given that active Go2α binds Rap1GAP, it may be speculated (contrary to the interaction of wild-type Go1α and Rap1GAP) that activated Go2α increases Rap1GAP activity. Whether this interaction with active Go2α leads to a decrease of phosphorylation or interferes with the Go1α mediated targeting to degradation remains unknown; however, both mechanisms lead to a decrease of Rap1 activity. Consequently, increased levels of active GTP-bound Rap1 in Go2α deletion mutants may be caused by an impairment of Rap1GAP activity. Thus, depending on their state of activity Go1α and Go2α may differentially control Rap1 activity and thereby influence Rap1 actions in neuronal development. Indeed axonal growth and branching is promoted in Go2α deletion mutants potentially through an increase of activated Rap1. An increase in Go1α levels in Go2α deletion mutants as explanation for higher amounts of Rap1GTP could be excluded, as previous quantification of Go1α expression in these mice showed no difference from wild type (Brunk et al. 2008).

Taken together, our data show that Go2α specifically regulates axonal outgrowth and branching. Interaction with Rap1GAP and a higher ratio of Rap1GTP/Rap1 in the Go2α deletion mutants indicate that Go2α may decrease Rap1 activity. Considering that Go1α, in its resting state, has been shown to activate Rap1 by promoting degradation of Rap1GAP the two splice variants of Goα may differentially regulate Rap1 activity and neurite outgrowth.


The authors are indebted to Birgit Metze, Marion Möbes, and Antje Dräger for skillfull technical assistance, to Sam Booker for linguistic corrections and to the Deutsche Forschungsgemeinschaft for financial support (DFG Ah67/3-3). Part of this research was supported by the Intramural Research Program of the NIH (Project Z01-ES-101643 to LB). None of the authors has to declare any conflict of interest.