The Sushi Domains of GABAB Receptors Function as Axonal Targeting Signals

GABAB receptors are the G-protein-coupled receptors for GABA, the main inhibitory neurotransmitter in the brain. Two receptor subtypes, GABAB(1a,2) and GABAB(1b,2), are formed by the assembly of GABAB1a and GABAB1b subunits with GABAB2 subunits. The GABAB1b subunit is a shorter isoform of the GABAB1a subunit lacking two N-terminal protein interaction motifs, the sushi domains. Selectively GABAB1a protein traffics into the axons of glutamatergic neurons, whereas both the GABAB1a and GABAB1b proteins traffic into the dendrites. The mechanism(s) and targeting signal(s) responsible for the selective trafficking of GABAB1a protein into axons are unknown. Here, we provide evidence that the sushi domains are axonal targeting signals that redirect GABAB1a protein from its default dendritic localization to axons. Specifically, we show that mutations in the sushi domains preventing protein interactions preclude axonal localization of GABAB1a. When fused to CD8α, the sushi domains polarize this uniformly distributed protein to axons. Likewise, when fused to mGluR1a the sushi domains redirect this somatodendritic protein to axons, showing that the sushi domains can override dendritic targeting information in a heterologous protein. Cell surface expression of the sushi domains is not required for axonal localization of GABAB1a. Altogether, our findings are consistent with the sushi domains functioning as axonal targeting signals by interacting with axonally bound proteins along intracellular sorting pathways. Our data provide a mechanistic explanation for the selective trafficking of GABAB(1a,2) receptors into axons while at the same time identifying a well defined axonal delivery module that can be used as an experimental tool.


Introduction
GABA B receptors exert distinct regulatory effects on synaptic transmission (Couve et al., 2000;Bowery et al., 2002;. Presynaptic GABA B receptors inhibit the release of GABA (autoreceptors) and other neurotransmitters (heteroreceptors), while postsynaptic GABA B receptors inhibit neuronal excitability by activating K ϩ channels. Receptor subtypes are based on the subunit isoforms GABA B1a and GABA B1b , both of which combine with GABA B2 subunits to form two heteromeric receptors, GABA B(1a,2) and GABA B(1b,2) (Marshall et al., 1999). Most if not all neurons in the CNS coexpress GABA B(1a,2) and GABA B(1b,2) receptors. The GABA B1a and GABA B1b subunit isoforms derive from the same gene by alternative promoter usage and solely differ in their N-terminal ectodomains (Kaupmann et al., 1997;Steiger et al., 2004). GABA B1a contains at its N terminus two sushi domains (SDs) that are lacking in GABA B1b (Hawrot et al., 1998). SDs, also known as complement control protein (CCP) modules or short consensus repeats (SCR), are conserved protein interaction motifs present in proteins of the complement system, in adhesion molecules and in G-protein-coupled receptors (Morley and Campbell, 1984;Kirkitadze and Barlow, 2001;Grace et al., 2004;Lehtinen et al., 2004;Perrin et al., 2006). The tertiary structure of SDs is fixed by two intramolecular disulfide bridges that are critical for interaction with other proteins (Soares and Barlow, 2005). Consistent with their role as interaction motifs, the SDs of GABA B1a recognize binding sites in neuronal membranes (Tiao et al., 2008).
The individual functions of the GABA B1a and GABA B1b subunit isoforms were dissected by comparing genetically modified 1a Ϫ/ Ϫ and 1b Ϫ/ Ϫ mice, which express either one or the other isoform (Pérez-Garci et al., 2006;Shaban et al., 2006;Vigot et al., 2006;Guetg et al., 2009). It was found that only GABA B(1a,2) receptors inhibit glutamate release in response to endogenous GABA, while both GABA B(1a,2) and GABA B(1b,2) receptors mediate postsynaptic inhibition. This is a consequence of a selective trafficking of GABA B(1a,2) receptors into axons. Specifically, experiments with organotypic slice cultures revealed that heterologously expressed GABA B1a subunits traffic to axons and dendrites, while GABA B1b subunits traffic to dendrites only (Vigot et al., 2006). The signals and mechanisms leading to a somatodendritic expression of GABA B1b subunits and a more uniform distribution of GABA B1a subunits are unknown. In general, polarized sorting of transmembrane proteins relies on signals in the targeted protein themselves (Craig and Banker, 1994;Winckler and Mellman, 1999). Since the targeting location of the shorter GABA B(1b,2) receptor is the somatodendritic compartment, this suggests that the longer GABA B(1a,2) receptor also contains common dendritic targeting signals in either the GABA B1a or the associated GABA B2 subunit. This implies a mechanism that prevents a fraction of GABA B(1a,2) receptors from trafficking to the default somatodendritic compartment and instead directs them to axons.
Here, we report that GABA B(1a,2) receptors are trafficked into axons by the SDs, which function as axonal targeting signals along intracellular sorting pathways. We discuss the mechanistic and regulatory implications of our findings.

Materials and Methods
Mouse strains. Primary neuronal cultures were prepared from WT BALB/c mice or 1a Ϫ/ Ϫ , 1b Ϫ/ Ϫ , and 2 Ϫ/ Ϫ mice that were strictly kept in the BALB/c inbred background Gassmann et al., 2004;Vigot et al., 2006). All animal experiments were subjected to institutional review and conducted in accordance with Swiss guidelines and approved by the veterinary office of Basel-Stadt.
Generation of mutant proteins. Cloning of Myc-tagged expression constructs was based on a strategy described earlier (Pagano et al., 2001). Briefly, to allow detection of transiently expressed subunits, the intrinsic signal peptides were replaced by 36 residues encoding the mGluR5 signal peptide MVLLLILSVLLLKEDVRGSAQS, followed by the Myc-tag, TREQKLISEEDLTR [replaced residues: Myc-GB1a, 1-16 (Kaupmann et al., 1997); Myc-GB1b, 1-29 (Kaupmann et al., 1997); Myc-mGluR1a, 1-20 (Masu et al., 1991); Myc-CD8␣, 1-21]. The mGluR5 signal peptide was used because it is known to accurately release N-terminal epitope tags (Ango et al., 1999). To generate Myc-GB1aCS, the four cysteine residues of GABA B1a at positions 29, 95, 99, and 156 (Kaupmann et al., 1997) were mutated to serine residues by site-directed mutagenesis of thymine to adenine. To generate Myc-GB1a⌬SD1 and Myc-GB1a⌬SD2, residues G 28 to C 95 or V 96 to Q 157 of Myc-GB1a were deleted. To generate Myc-SDs-mGluR1a, residues G 17 to S 134 of GABA B1a were introduced after the Myc-tag in rat Myc-mGluR1a (mGluR1a was a gift from R. M. Duvoisin, Oregon Health and Science University, Portland, OR). To generate Myc-SDs-CD8␣, the residues G 17 to S 134 of GABA B1a were introduced after the Myc-tag in Myc-CD8␣ (CD8␣ was a gift from G. A. Banker, Oregon Health and Science University, Portland, OR). Initially, all constructs were subcloned into the cytomegalovirus-based eukaryotic expression vector pCI (Promega) to confirm protein expression in HEK293 cells. Subsequently all constructs were shuttled into plasmid pMH4-SYN-1 for expression under control of the synapsin-1 promoter in cultured hippocampal neurons ͓gift from T. G. Oertner (Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland) and K. Svoboda (Howard Hughes Medical Institute, Ashburn, VA)͔. In GB1a-GFP and GB1b-GFP, the coding sequence for GFP was cloned in frame at the C terminus of full-length GABA B1a and GABA B1b (Kaupmann et al., 1997), leaving the cognate signal peptides unaltered. All constructs were verified by sequencing.
Neuronal culture and transfection. Cultured hippocampal neurons were prepared as described previously (Brewer et al., 1993;Goslin et al., 1998). Briefly, embryonic day 16.5 mouse hippocampi were dissected, digested with 0.25% trypsin in Hank's solution (Invitrogen) for 15 min at 37°C, dissociated by trituration, and plated on glass coverslips coated with 1 mg/ml poly-L-lysine hydrobromide (Sigma) in 0.1 M borate buffer (boric acid/sodium tetraborate). Neurons were seeded at low density (ϳ100 -150 cells/mm 2 ) for endogenous GABA B1 labeling or at high density (ϳ750 cells/mm 2 ) for transfection experiments or electrophysiological recordings and then incubated at 37°C/5% CO 2 . Low-density cultures were cultivated in HC-MEM medium [1ϫ MEM with Glutamax, 0.3% glucose (w/v), 10% horse serum, and 1% Pen/Strep] for the first 4 h to allow neurons to attach. Subsequently, the coverslips were transferred to a feeder layer of primary astrocytes in serum-free medium [1ϫ MEM with Glutamax, 0.3% glucose (w/v), and 1% Pen/Strep] sup-plemented with 1% N2 (Invitrogen). Primary astrocytes were obtained from newborn P0-P1 BALB/c mice. To prevent extensive proliferation of astrocytes 5 M arabinoside (AraC, Sigma) was added to the culture medium after 2 d. High-density cultures were grown in Neurobasal medium supplemented with B27 (Invitrogen), 0.5 mM L-glutamine, and 50 -100 g/ml Pen/Strep. In addition, 25 M glutamic acid was added to the medium for the first 3 d. At DIV5, neurons were cotransfected with the appropriate expression constructs and soluble RFP (pMH4-SYN-tdimer2-RFP; gift from R. Tsien, University of California San Diego, La Jolla, CA) using Lipofectamine 2000 transfection reagent (Invitrogen).
Electrophysiology. Hippocampal neurons were cultured for 2-3 weeks. On the day of the experiment, coverslips were placed in an interface chamber containing saline solution (140 mM NaCl, 3 mM KCl, 2.5 mM CaCl 2 , 1.2 mM MgCl 2 , 11.1 mM glucose, 10 mM HEPES, pH 7.2) equilibrated with 95% O 2 /5% CO 2 at 30 -32°C. Neurons were visualized using infrared and differential interference contrast optics. Whole-cell patchclamp recordings were performed at Ϫ60 mV from the somata of neurons to measure mEPSCs in the presence of tetrodotoxin (1 M) and bicuculline (10 M). Patch electrodes (ϳ3 M⍀) were filled with a solution containing the following: 140 mM Cs-gluconate, 10 mM HEPES, 10 mM phosphocreatine, 5 mM QX-314, 4 mM Mg-ATP, 0.3 mM Na-GTP, at pH 7.2 with Cs-OH and 285 mOsm. During the experiment drugs were applied by superfusion into the recording chamber. GABA B receptors were activated by baclofen (100 M) and inactivated by the selective antagonist CGP54626 (1 M). Detection and analysis of mEPSCs was performed by MiniAnalysis software (version 6.0.4, Synaptosoft). Experiments with CHO cells expressing WT or mutant GABA B receptors together with Kir3.1/3.2 channels and EGFP (used as a transfection marker) were performed at room temperature (RT) 2 d after transfection with Lipofectamine 2000 (Invitrogen). As a negative control, CHO cells expressing Kir3.1/3.2 channels and EGFP in the absence of GABA B receptors were used. Cells were continuously superfused with an extracellular solution composed of the following (in mM): 145 NaCl, 2.5 KCl, 1 MgCl 2 , 2 CaCl 2 , 10 HEPES, 25 glucose; pH 7.3, 323 mOsm. Patch pipettes were filled with an intracellular solution composed of the following (in mM): 107.5 potassium gluconate, 32.5 KCl, 10 HEPES, 5 EGTA, 4 MgATP, 0.6 NaGTP, 10 Tris phosphocreatine; pH 7.2, 297 mOsm. GABA B responses were evoked by application of baclofen (10 s) (Dittert et al., 2006) and recorded with an Axopatch 200B patch-clamp amplifier. The presence of Kir3.1/3.2 channels in transfected cells was confirmed in voltage ramps from Ϫ150 mV to ϩ30 mV in the presence of a high extracellular potassium concentration (40 mM).
Microscopy. Immunolabeled neurons were viewed at room temperature on a Leica DM5000B fluorescence microscope. Glutamatergic neurons were discriminated from GABAergic neurons by their extensively branched spiny dendrites visualized by the RFP filling (Benson et al., 1994;Obermair et al., 2003). Digital pictures were captured using Soft Imaging System and AnalySIS software (F-View) and identically processed with Adobe Photoshop (RGB input levels, brightness/contrast). The filters used to detect secondary antibodies were as follows: L5-filter for Alexa goat anti-mouse 488 (Myc antibody), Y5-filter for Alexa goat anti-chicken 647 (MAP2 antibody), Y3-filter for RFP or Alexa goat antirabbit 568 (polyclonal GABA B1 antibody). The ER-targeted GFP was from Clontech. Pictures were taken with each filter separately. Pictures from the endogenous GABA B1 staining were captured using immersion oil without autofluorescence (Leica Microsystems catalog #11513859) and a 63ϫ oil objective with 1.32 NA (HCX PL APO). Images to evaluate the axonal versus dendritic distribution of heterologously expressed GABA B protein were captured using a 20ϫ air objective 0.7 NA (HC PLAN APO).
Quantification of axonal versus dendritic distribution. The axon-todendrite (A:D) ratio of endogenous GABA B1 protein was determined using MetaMorph Imaging software. One-pixel-wide lines were traced along representative axons and dendrites in the tubulin-stained images. Next to each line, a rectangle was drawn for background subtraction. Subsequently, the lines and rectangles were transferred to the corresponding pic-ture with GABA B1 immunostaining. Average pixel intensities were determined along the traced lines and in the background rectangles. After background subtraction, the anti-GABA B1 fluorescence intensity was normalized to the anti-tubulin fluorescence intensity in axon and dendrites. The normalized data were used to determine the A:D ratio. The A:D ratio of Myc-tagged constructs was determined by normalizing the Myc-labeling to the RFP labeling (Gu et al., 2003;Sampo et al., 2003). Cells to be analyzed were selected using the soluble RFP fill and only considered for quantification if the RFP fluorescence was evenly distributed over the entire neuron, including distal axons and dendrites. Cells expressing constructs at very high levels were excluded from analysis because such cells exhibit a less polarized distribution of expressed proteins. Seven to nineteen neurons from at least two independent culture preparations for each construct were analyzed. SPSS or GraphPad PRISM software was used for statistical analysis.

Results
Endogenous GABA B1a but not GABA B1b subunits inhibit glutamate release and localize to axons in cultured hippocampal neurons Pyramidal neurons typically make up 85-90% of neurons in dissociated hippocampal cultures (Goslin et al., 1998) and potentially provide a simple experimental system to study the targeting of transfected GABA B1a and GABA B1b subunits in glutamatergic neurons. We first investigated whether cultured pyramidal neurons preserve the selective association of GABA B(1a,2) receptors with glutamatergic terminals seen in hippocampal slices (Vigot et al., 2006;Guetg et al., 2009). Specifically, we addressed whether functional GABA B heteroreceptors are present in cultured pyramidal neurons of 1b Ϫ/ Ϫ mice, but absent in neurons of 1a Ϫ/ Ϫ mice. Activation of GABA B heteroreceptors by baclofen, a GABA B receptor agonist, inhibits the spontaneous release of glutamate and as a result reduces the miniature EPSC (mEPSC) frequency (Yamada et al., 1999;Tiao et al., 2008). We found that baclofen strongly reduced the mEPSC frequencies in wild-type (WT) and 1b Ϫ/ Ϫ neurons, while baclofen only marginally reduced the mEPSC frequency in 1a Ϫ/ Ϫ neurons ( Fig.  1 A, B). This confirms that functional GABA B heteroreceptors are specifically lacking in cultured hippocampal neurons of 1a Ϫ/ Ϫ mice. Weak residual heteroreceptor activity in 1a Ϫ/ Ϫ mice in response to high concentrations of baclofen was also observed in acute hippocampal slices (Vigot et al., 2006;Guetg et al., 2009). This may reflect that low amounts of GABA B(1b,2) receptors are present at glutamatergic terminals. Alternatively, baclofen B, Summary bar graph illustrating that baclofen strongly inhibits the frequency of mEPSCs in WT (78.1 Ϯ 3.1%, n ϭ 16) and 1b Ϫ/ Ϫ (70.8 Ϯ 5.1%, n ϭ 15) neurons, but not in 1a Ϫ/ Ϫ (7.7 Ϯ 2.8%, n ϭ 10) neurons. Values are means Ϯ SEM, one-sample t test, *p Ͻ 0.05, ***p Ͻ 0.001. C, Cultured hippocampal neurons from WT, 1a Ϫ/ Ϫ , and 1b Ϫ/ Ϫ mice were fixed, permeabilized, and stained with antibodies recognizing GABA B1a and GABA B1b (GB1), the dendritic marker protein MAP2, or the cytoskeleton protein tubulin. Arrows mark MAP2-negative axons. Note the lack of GB1 immunolabeling in axons of 1a Ϫ/ Ϫ neurons. Scale bar, 50 m. D, A:D ratio of the endogenous GABA B1 proteins in WT, 1a Ϫ/ Ϫ , and 1b Ϫ/ Ϫ neurons. The fluorescence intensity of GB1 immunolabeling was normalized to the fluorescence intensity of tubulin immunolabeling. The A:D ratio of GABA B1 protein is significantly smaller in 1a Ϫ/ Ϫ compared to WT and 1b Ϫ/ Ϫ neurons (mean Ϯ SEM, ***p Ͻ 0.001, 1-way ANOVA with Tukey's post hoc test). E, Schematic depiction of endogenous GABA B(1a,2) and GABA B(1b,2) receptor distribution in cultured hippocampal neurons and hippocampal slice culture. Squares indicate the two in tandem arranged SDs at the N terminus of GABA B1a . may also activate somatic GABA B(1b,2) receptors and the ensuing hyperpolarizing potentials passively propagate to glutamatergic terminals, where they contribute to presynaptic inhibition (Alle and Geiger, 2006). We next analyzed the expression levels of the endogenous GABA B1a and GABA B1b proteins in axons and dendrites of cultured hippocampal neurons. Due to the lack of GABA B1a -or GABA B1b -specific antibodies, we used cultured hippocampal neurons from 1a Ϫ/ Ϫ and 1b Ϫ/ Ϫ mice and stained them with an antibody recognizing the common C-term of GABA B1 subunits (Kulik et al., 2002). To distinguish dendrites from axons, we immunolabeled the dendritic microtubule-associated protein MAP2 and tubulin, a constituent of axons and dendrites (Caceres et al., 1984). In WT and 1b Ϫ/ Ϫ pyramidal neurons, GABA B1 immunostaining was observed in MAP2-positive somata and dendrites as well as in MAP2-negative axons (Fig. 1C). In contrast, in cultured 1a Ϫ/ Ϫ pyramidal neurons, GABA B1 immunostaining was restricted to the somatodendritic compartment. This confirms that primarily GABA B1a localizes to axons in cultured pyramidal neurons. To determine the axon-to-dendrite (A:D) ratio of the endogenous GABA B1 proteins, we normalized the red fluorescence intensity of the GABA B1 staining to the green fluorescence intensity of the tubulin staining in axons and dendrites. In all three genotypes the A:D ratio was Ͻ1, indicating that most GABA B1 protein is localized somatodendritically (WT: 0.54 Ϯ 0.05, n ϭ 7; 1a Ϫ/ Ϫ : 0.22 Ϯ 0.02, n ϭ 7; 1b Ϫ/ Ϫ : 0.60 Ϯ 0.05, n ϭ 8; p Ͻ 0.001 for 1a Ϫ/ Ϫ vs WT and 1b Ϫ/ Ϫ ). However, the A:D ratio in 1a Ϫ/ Ϫ neurons was significantly reduced compared to WT and 1b Ϫ/ Ϫ neurons (Fig. 1D), indicating that significantly more GABA B1a than GABA B1b protein enters the axonal compartment. In summary, our electrophysiological and immunocytochemical analysis demonstrates that cultured pyramidal neurons preserve the preferential association of GABA B1a with glutamatergic terminals seen in hippocampal slices (Fig. 1 E).

Exogenous GABA B1a and GABA B1b subunits reproduce the distribution patterns of the endogenous subunits
We next assessed whether GABA B1 isoforms with an N-terminal Myc-tag (Myc-GB1a, Myc-GB1b) recapitulate the subcellular distribution of the endogenous proteins when expressed in cultured hippocampal neurons. Cultured hippocampal neurons were transfected after 5 d in vitro (DIV5) with Myc-GB1a or Myc-GB1b cDNAs under control of the neuron-specific synapsin-1 promoter (Kügler et al., 2001;Boulos et al., 2006), as this promoter avoids randomization of distribution patterns due to overexpression (Vigot et al., 2006). To accurately release the N-terminal Myc-epitope in the Myc-GB1a and Myc-GB1b proteins, we used a surrogate signal peptide instead of the intrinsic signal peptides (Ango et al., 1999). We coexpressed Myc-GB1a or Myc-GB1b with the freely diffusible red fluorescent protein (RFP) tdimer2, which outlines the morphology of the transfected neurons. Following transfection, neurons were fixed at DIV14, permeabilized, and stained with antibodies against the Myc-tag and the dendritic marker MAP2. We found that Myc-GB1a was present in axons, somata, and dendrites, whereas Myc-GB1b was restricted to the somatodendritic compartment (Fig.  2 A). The A:D ratios of Myc-GB1a and Myc-GB1b were determined by normalizing the green Myc fluorescence intensity to the RFP fluorescence intensity in axons and dendrites (Gu et al., 2003;Sampo et al., 2003;Das and Banker, 2006). The A:D ratio for transfected Myc-GB1a was increased by 2.7-fold compared to Myc-GB1b (Myc-GB1a: 0.38 Ϯ 0.04, n ϭ 10; Myc-GB1b: 0.14 Ϯ 0.05, n ϭ 10; p Ͻ 0.01) (Fig. 2 B), analogous as with the endogenous GABA B1a and GABA B1b proteins in 1b Ϫ/ Ϫ and 1a Ϫ/ Ϫ neurons, respectively (Fig. 1 D). This demonstrates that the trafficking of endogenous and transfected GABA B1 subunits is alike. Moreover, this indicates that neither putative compensatory mechanisms in the knock-out backgrounds nor the surrogate signal peptide interfere with trafficking. We nevertheless also determined the distribution patterns of GABA B1 proteins that are C-terminally tagged with the green fluorescent protein (GFP) and therefore contain their intrinsic signal peptides. The A:D ratio for GB1a-GFP was significantly increased by twofold compared to GB1b-GFP (GB1a-GFP: 0.49 Ϯ 0.06, n ϭ 7; GB1b-GFP: 0.25 Ϯ 0.04, n ϭ 7; p Ͻ 0.01), thus consolidating that the surrogate signal peptide and the intrinsic signal peptides lead to a comparable axonal versus dendritic distribution. Furthermore, we analyzed whether trafficking is influenced by the developmental stage of cultured neurons. In neurons at DIV21, the A:D ratio of Myc-GB1a was significantly increased compared to Myc-GB1b (Myc-GB1a: 0.49 Ϯ 0.04, n ϭ 6; Myc-GB1b: 0.25 Ϯ 0.05, n ϭ 6; p Ͻ 0.01) (supplemental Fig. S1, available at www.jneurosci. org as supplemental material), providing no evidence for a developmental regulation of trafficking.
The levels of Myc-GB1a and Myc-GB1b at the cell surface were too low for reliable quantification. Presumably, exogenous GABA B1 subunits compete with endogenous GABA B1 subunits for GABA B2 , which is required for escorting GABA B1 to the plasma membrane (Margeta-Mitrovic et al., 2000;Pagano et al., 2001). To increase surface expression levels of the exogenous GABA B1 proteins, we therefore coexpressed the GABA B2 protein with the individual Myc-GB1a and Myc-GB1b proteins. This allowed quantification of the Myc-fluorescence at the cell surface of nonpermeabilized cells. The Mycfluorescence was normalized to the fluorescence of coexpressed RFP and the A:D ratio determined as described above. Surface Myc-GB1a exhibited a significantly increased A:D ratio compared to surface Myc-GB1b (Myc-GB1a: 0.50 Ϯ 0.09, n ϭ 10; Myc-GB1b: 0.26 Ϯ 0.02, n ϭ 10; p Ͻ 0.05) (Fig.  2 B), demonstrating that GABA B1a is also enriched over GABA B1b at the axonal plasma membrane. In addition, comparison of the data in Figure 2 B shows that significantly more GABA B1a than GABA B1b protein traffics to axons, regardless of whether or not exogenous GABA B2 is supplied to WT neurons. This demonstrates that the GABA B2 expression level does not markedly influence the axonal versus dendritic distribution of the GABA B1a and GABA B1b proteins.

Each SD in GABA B1a can mediate axonal localization on its own
The SDs in GABA B1a bind with low nanomolar affinity to binding sites in neuronal membranes (Tiao et al., 2008) and likely mediate axonal localization through interaction with other protein(s). To interact with binding partners the SDs in GABA B1a need to fold into a globular structure that is stabilized by disulfide bonds (Wei et al., 2001;Tiao et al., 2008).
We therefore addressed whether the tertiary structure of the SDs is crucial for axonal localization of GABA B1a . In the Myc-GB1aCS mutant, we prevented disulfide bond formation in each of the SDs by converting two of the four conserved cysteines into serines. Following transfection into cultured hippocampal neurons, Myc-GB1aCS was robustly targeted to dendrites but not to axons (Fig. 3A). Accordingly, the A:D ratio in Myc-GB1aCS was significantly smaller than that for WT Myc-GB1a (Myc-GB1a: 0.41 Ϯ 0.06, n ϭ 8; Myc-GB1aCS: 0.14 Ϯ 0.03, n ϭ 10; p Ͻ 0.001) (Fig. 3C). Of note, the A:D ratio of Myc-GB1aCS was similar to that of Myc-GB1b (Fig. 2 B). While Myc-GB1aCS failed to traffic to axons the mutant protein efficiently activated Kir3 channels when coexpressed with GABA B2 (Fig. 3D). This demonstrates that interfering with the folding of the SDs impairs axonal trafficking without impairing receptor surface expression or G-protein signaling. Altogether, these results support that the SDs engage in interactions that are necessary for axonal localization of GABA B1a . Structurally, the two SDs in GABA B1a differ from each other (Blein et al., 2004). The first SD shows conformational heterogeneity under a wide range of conditions and interacts with the extracellular matrix protein fibulin-2. The second SD is more compactly folded and exhibits strong structural similarity with the SDs in proteins of the complement system. It is conceivable that the two SDs exert different functions and interact with different proteins. We therefore investigated whether each of the two SDs in GABA B1a can mediate axonal targeting on its own. In the Myc-GB1a⌬SD1 and Myc-GB1a⌬SD2 mutants, we deleted either the first or the second SD, respectively (Fig. 3B). Myc-GB1a⌬SD1 and Myc-GB1a⌬SD2 were both efficiently targeted to axons, and the A:D ratios were not significantly different from that of WT Myc-GB1a (Myc-GB1a: 0.41 Ϯ 0.06, n ϭ 8; Myc-GB1a⌬SD1: 0.49 Ϯ 0.04, n ϭ 9; Myc-GB1a⌬SD2: 0.47 Ϯ 0.04, n ϭ 8; p Ͼ 0.05) (Fig. 3C). This shows that each of the two SDs in GABA B1a can mediate axonal localization on its own. The SDs of GABA B1a polarize the uniformly distributed transmembrane protein CD8␣ to axons The SDs could promote axonal localization of GABA B1a either by acting as axonal trafficking signals or, alternatively, by inactivating dendritic targeting signals, which would also result in a more uniform distribution. To distinguish between these two possibilities, we analyzed whether the SDs of GABA B1a are capable of polarizing an unpolarized heterologous transmembrane protein, CD8␣ (Jareb and Banker, 1998), to axons. We first confirmed that Myc-CD8␣ uniformly distributes to axons and dendrites of transfected hippocampal neurons (Fig. 4). As expected for an unpolarized protein, the A:D ratio was with 1.24 Ϯ 0.07 (n ϭ 24) close to 1. In contrast, when the two SDs of GABA B1a were fused to the ectodomain of CD8␣, the chimeric Myc-SDs-CD8␣ protein clearly polarized to axons (A:D ratio 2.37 Ϯ 0.26, n ϭ 24; p Ͻ 0.001 vs Myc-CD8␣) (Fig. 4). This clearly identifies the SDs as bona fide axonal targeting signals.
The SDs of GABA B1a direct the somatodendritic mGluR1a protein to axons According to our hypothesis, the SDs of GABA B1a not only act as axonal trafficking signals but also override the dendritic targeting signals present in GABA B1a and/or GABA B2 . We therefore investigated whether the SDs of GABA B1a can direct a somatodendritically localized heterologous transmembrane protein to axons. For this experiment, we used mGluR1a, a receptor with C-terminal dendritic trafficking signals (Francesconi and Duvoisin, 2002;Das and Banker, 2006). We confirmed that Myc-mGluR1a is highly expressed in the dendrites but ex-cluded from the axons of transfected hippocampal neurons (Fig. 5 A, B). When the two SDs of GABA B1a were fused to the N-terminal ectodomain of mGluR1a, the chimeric Myc-SDs-mGluR1a protein readily trafficked to axons and exhibited a significantly higher A:D ratio than WT Myc-mGluR1a (Myc-mGluR1a: 0.03 Ϯ 0.06, n ϭ 9; Myc-SDs-mGluR1a: 1.26 Ϯ 0.15, n ϭ 11; p Ͻ 0.001). This shows that the SDs of GABA B1a can override the somatodendritic targeting signals in the C terminus of mGluR1a.

Surface expression is not required for axonal delivery of GABA B1a
GABA B1a is not only present in the axons, but also highly expressed in the somatodendritic compartment (Figs. 1, 2). It is therefore conceivable that GABA B1a reaches the axonal compartment through transcytosis from the somatodendritic compartment, similar to what is reported for the neuronal cell adhesion molecule NgCAM (Wisco et al., 2003). This dendrite-toaxon transcytotic pathway requires internalization of axonally bound proteins from the dendritic plasma membrane. We investigated whether Myc-GB1a can be transported into axons in the absence of surface expression. Since GABA B2 is necessary for surface localization of GABA B1 subunits (Margeta-Mitrovic et al., 2000;Pagano et al., 2001), we prevented surface trafficking of Myc-GB1a by expressing it in cultured hippocampal neurons of GABA B2 Ϫ/ Ϫ (2 Ϫ/ Ϫ ) mice (Gassmann et al., 2004). Myc-GB1a was transported into axons in the absence of GABA B2 (Fig. 6) and the A:D ratio in 2 Ϫ/ Ϫ neurons was not significantly different from that in WT neurons (Myc-GB1a in WT: 0.45 Ϯ 0.05, n ϭ 12; Myc-GB1a in 2 Ϫ/ Ϫ : 0.40 Ϯ 0.04, n ϭ 19; p Ͼ 0.05). This corroborates that Myc-GB1a reaches the axonal compartment via an intracellular route, independent of any surface expression. Lateral diffusion of surface receptors is therefore not necessary for axonal localization of GABA B1a . However, the SDs are not only involved in axonal delivery of GABA B receptors but also in their retention at the cell surface of the terminal (Tiao et al., 2008). Lateral diffusion and selective retention could therefore, in principle, contribute to the pool of axonal GABA B1a receptors. It was recently proposed that proteins not only traffic into axons via post-Golgi transport vesicles but also within the endoplasmic reticulum (ER), from where proteins are released via exit sites (Aridor and Fish, 2009;Merianda et al., 2009). It is therefore conceivable that GABA B1a traffics into axons within the ER. As previously reported (Ramírez et al., 2009), we found a partial colocalization of transfected GABA B1a subunits with the ER in the somatodendritic compartment using an ER-targeted GFP (Aoki et al., 2002) as a marker (supplemental Fig. S2, available at www. jneurosci.org as supplemental material). We also observed a partial colocalization of transfected GABA B1a with ER-targeted GFP in axons, making it conceivable that some GABA B1a also enters axonal ER. However, according to prevailing concepts axonally destined proteins traffic in intracellular post-Golgi transport vesicles to the terminals (Horton and Ehlers, 2003). We therefore expect that intracellular GABA B1a in axons is mostly present in transport vesicles delivering their cargo to the terminal.