Dendritic Assembly of Heteromeric γ-Aminobutyric Acid Type B Receptor Subunits in Hippocampal Neurons

Understanding the mechanisms that control synaptic efficacy through the availability of neurotransmitter receptors depends on uncovering their specific intracellular trafficking routes. γ-Aminobutyric acid type B (GABAB) receptors (GABABRs) are obligatory heteromers present at dendritic excitatory and inhibitory postsynaptic sites. It is unknown whether synthesis and assembly of GABABRs occur in the somatic endoplasmic reticulum (ER) followed by vesicular transport to dendrites or whether somatic synthesis is followed by independent transport of the subunits for assembly and ER export throughout the somatodendritic compartment. To discriminate between these possibilities we studied the association of GABABR subunits in dendrites of hippocampal neurons combining live fluorescence microscopy, biochemistry, quantitative colocalization, and bimolecular fluorescent complementation. We demonstrate that GABABR subunits are segregated and differentially mobile in dendritic intracellular compartments and that a high proportion of non-associated intracellular subunits exist in the brain. Assembled heteromers are preferentially located at the plasma membrane, but blockade of ER exit results in their intracellular accumulation in the cell body and dendrites. We propose that GABABR subunits assemble in the ER and are exported from the ER throughout the neuron prior to insertion at the plasma membrane. Our results are consistent with a bulk flow of segregated subunits through the ER and rule out a post-Golgi vesicular transport of preassembled GABABRs.

Understanding the mechanisms that control synaptic efficacy through the availability of neurotransmitter receptors depends on uncovering their specific intracellular trafficking routes. ␥-Aminobutyric acid type B (GABA B ) receptors (GABA B Rs) are obligatory heteromers present at dendritic excitatory and inhibitory postsynaptic sites. It is unknown whether synthesis and assembly of GABA B Rs occur in the somatic endoplasmic reticulum (ER) followed by vesicular transport to dendrites or whether somatic synthesis is followed by independent transport of the subunits for assembly and ER export throughout the somatodendritic compartment. To discriminate between these possibilities we studied the association of GABA B R subunits in dendrites of hippocampal neurons combining live fluorescence microscopy, biochemistry, quantitative colocalization, and bimolecular fluorescent complementation. We demonstrate that GABA B R subunits are segregated and differentially mobile in dendritic intracellular compartments and that a high proportion of non-associated intracellular subunits exist in the brain. Assembled heteromers are preferentially located at the plasma membrane, but blockade of ER exit results in their intracellular accumulation in the cell body and dendrites. We propose that GABA B R subunits assemble in the ER and are exported from the ER throughout the neuron prior to insertion at the plasma membrane. Our results are consistent with a bulk flow of segregated subunits through the ER and rule out a post-Golgi vesicular transport of preassembled GABA B Rs.
The efficacy of synaptic transmission depends on the intracellular trafficking of neurotransmitter receptors (1,2). The trafficking of glutamatergic and GABA A 6 receptors has been extensively studied, and their implications for synaptic plasticity have been well documented (3,4). For example, differential trafficking of ␣-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors modifies synaptic strength and influences experience-dependent plasticity in vivo (5). The molecular mechanisms that govern the trafficking of metabotropic GABA B Rs and their consequences for synaptic inhibition remain less clear. In particular, limited information is available regarding the relationship between the trafficking of GABA B Rs and the topological complexity of the secretory pathway in neurons.
GABA B Rs mediate the slow component of synaptic inhibition by acting on pre-and postsynaptic targets (6 -8). They are implicated in epilepsy, anxiety, stress, sleep disorders, nociception, depression, and cognition (9). They also represent attractive targets for the treatment of withdrawal symptoms from drugs of addiction such as cocaine (10). They are obligatory heteromers composed of GABA B R1 and GABA B R2 subunits. GABA B R1 contains an RXR-type sequence in the intracellular C-terminal domain that functions as an ER retention motif (11,12). The ER retention sequence is masked upon assembly with GABA B R2 resulting in the appearance of functional receptors at the plasma membrane. Only GABA B R1 binds GABA with high affinity, whereas G protein signaling is exclusively mediated by the second and third intracellular loops of GABA B R2 (13)(14)(15). GABA B Rs are located in dendrites and axons, but their distribution does not coincide with the active zone or the postsynaptic density. Rather, they are adjacent to both compartments constituting perisynaptic receptors (16,17).
If GABA B R subunits are synthesized in the soma, at least two possibilities exist for their anterograde transport, assembly, and insertion in dendrites. First, the subunits may be synthesized in the cell body, assembled in the somatic ER, and targeted preassembled in post-Golgi vesicles to their site of insertion in dendrites. Alternatively, they may be synthesized in the soma and transported through the ER membrane as non-heteromeric subunits. In the latter scenario, newly assembled receptors may exit the ER throughout the somatodendritic compartment prior to insertion at the plasma membrane and diffuse laterally for retention at functional sites. No evidence exists to discriminate between these possibilities. We reasoned that a prevalence of associated subunits in post-Golgi vesicles in dendrites would favor the first alternative, whereas the existence of nonassociated subunits in intracellular compartments would support a somatodendritic assembly mechanism. Here we explore the presence of associated GABA B R subunits using fluorescence recovery after photobleaching (FRAP), biochemistry, and quantitative colocalization. In addition, we report for the first time the use of BiFC (18) to study GAB-A B R assembly in neurons. Our results demonstrate that GABA B R subunits are differentially mobile in dendrites and that a high proportion of non-associated subunits prevail in an intracellular fraction of the adult brain. They also show that GABA B R subunits are heteromeric at the plasma membrane but segregated in intracellular compartments of dendrites of hippocampal neurons. Importantly, treatment with brefeldin A (BFA) or interference of the coatomer protein complex II impair ER export and result in the accumulation of assembled subunits in intracellular compartments throughout the somatodendritic arbor. We conclude that GABA B R subunits are synthesized in the soma and remain segregated in intracellular compartments prior to somatodendritic assembly. Our observations rule out a post-Golgi vesicular transport of preassembled GABA B Rs and suggest an alternative mechanism of receptor targeting.

EXPERIMENTAL PROCEDURES
Animals-Adult pregnant female Sprague-Dawley rats were purchased from the Central Animal Facility at Universidad Católica de Chile and killed by asphyxia in a CO 2 chamber according to the Guide for Care and Use of Laboratory Animals (National Academy of Sciences, 1996).
Cell Lines, Neuronal Cultures, and Transfection-COS-7 cells were maintained and transfected as described previously using a GenePulser Xcell (Bio-Rad) (11). Primary hippocampal neurons were cultured from E18 rats and transfected by using Ca 2ϩ phosphate (19,20). Endogenous subunits interfere in the assembly of recombinant subunits only after 4 -5 days posttransfection; therefore, all experiments were analyzed between 48 and 72 h post-transfection.
Antibodies-Chicken GABA B R1 antibodies (which recognize GABA B R1a and GABA B R1b) and GABA B R2 antibodies against intracellular C-terminal domains were provided by S. J. Moss (Tufts University) and have been characterized previously (24). Guinea pig GABA B R1 (which recognize GABA B R1a and GABA B R1b), GABA B R2, and microtubuleassociated protein 2 antibodies were purchased from Chemicon (Temecula, CA). Piccolo and Bassoon antibodies were provided by E. D. Gundelfinger and W. D. Altrock (Leibniz Institute for Neurobiology, Magdeburg, Germany). Glutamate receptor (GluR) 2, cis-Golgi matrix protein 130 (GM130), and calnexin antibodies were purchased from BD Biosciences (San Jose, CA). KDEL antibodies (Grp78 and BiP) were purchased from Stressgen (Ann Arbor, MI). Myc antibodies were purchased from Sigma. Influenza A virus epitope (HA) antibodies were purchased from Roche Applied Science. Secondary anti-mouse, anti-rabbit, antiguinea pig, or anti-chicken antibodies conjugated to Texas Red, tetramethyl rhodamine isothiocyanate (TRITC), fluorescein isothiocyanate, or cyanine (Cy3) were purchased from Jackson ImmunoResearch Laboratories (West Grove, PA).
Image Capture, Colocalization, FRAP, and Time-lapse Microscopy-For colocalization and FRAP, confocal image stacks were captured with a Zeiss LSM-5, Pascal 5 Axiovert 200M microscope, using LSM 5 3.2 image capture and analysis software and a Plan-Apochromat 63ϫ/1.4 oil differential interference contrast objective. An approach based on Costes et al., using Manders coefficients (M 1 or M 2 ) was used to measure colocalization (28,29). All parameters were kept constant for a particular type of experiment (e.g. native GAB-A B R1 and GABA B R2 in dendrites). Images were acquired in 8 bits. Two-channel fluorescent image stacks (intensity, I(x,y,z) ʦ [0, 255]; voxel size, ⌬x/⌬y/⌬z ϭ 70/70/300 nm) were recorded in the multitrack mode. Channel-1 (fluorescein isothiocyanate) had an excitation/emission wavelength of exc / em ϭ 488/505-530 nm, and channel-2 (Texas Red/ TRITC) had an excitation/emission wavelength of with exc / em ϭ 543/Ͼ560 nm. We guaranteed that I(x,y,z) did not saturate and that image background was slightly above zero by carefully adjusting the laser power, detector gain, and detector offset. Digital zoom (4ϫ), rotation (none), and lines (four) were kept constant. In the multitrack mode, a spatial shift in the focal x-y plane was observed, probably due to mechanical factors such as the repositioning of the scanners. We calibrated the x-y shift with a grid and corrected all images before the calculation of colocalization coefficients. Raw confocal image stacks were deconvolved by Huygens Scripting software (Scientific Volume Imaging, Hilversum, Netherlands) using an algorithm based on the classic maximum likelihood estimator. Deconvolution improves the signal-to-noise ratio and is essential for a reliable analysis of colocalization coefficients. Image-processing routines were developed in our laboratory based on of Interactive Data Language (ITT, Boulder, CO), including routines for segmentation of different regions of interest (ROIs), visualization, calculation of colocalization coefficients, and statistical validation of the colocalization coefficients.
Reliable segmentation of ROIs in neuronal projections was achieved by applying gradient filters and selecting threshold values in the gradient histograms, which resulted in a homogeneous definition of the borders of receptor signals. Remaining holes inside receptor signals and segmented signals outside defined neuronal projections were filled or removed by morphological filters. The quality of the segmentation was controlled interactively by overlaying the original fluorescent images with the mask in each channel. Importantly, the segmentation criterion was kept constant for all images within a particular category (e.g. "dendrites").
For the quantification of colocalization we calculated M 1 and M 2 between ROIs 1 and 2, by the following definition, where M 1 and M 2 sum up the contribution of the respective fluorescence intensities in the colocalizing region I Ch1/2 (ROI1പROI2) and divide the number by the sum of the fluorescence intensities I Ch1/Ch2 inside of the segmented regions I Ch1 (ROI1) or I Ch2 (ROI2). ROI1 and ROI2 are segmented as described above and not by setting I(x,y,z) above zero in the respective fluorescent channels. M 1 and M 2 can be understood as the amount of colocalizing signal relative to the total amount of segmented signal. FRAP experiments were performed at ambient temperature in a 23°C equilibrated microscopy suite as described previously (25). Dendritic ROIs of 100 m 2 , 15 m away from the soma, were bleached for 8 s with the argon 488 nm laser at 100% power. The fluorescence recovery was measured every 10 s at 5% laser power by imaging the entire field. Boxes of 9 m 2 were selected within the original photobleached ROI to quantify recovery using ImageJ. For timelapse microscopy, images were obtained using an Olympus BX61WI upright microscope and an Olympus disk-scanning unit. Consecutive frames were acquired every 30 s over a period of 30 min.

GABA B R Subunits Are Transported to Dendrites and Display
Differential Mobility-To define the intracellular trafficking route used by GABA B R subunits we determined whether dendritic anterograde transport operated on protein subunits or whether mRNA targeting contributed to receptor localization. To discriminate between these possibilities the subcellular localization of GABA B R mRNAs in hippocampal neurons was examined via FISH. Importantly, both mRNAs concentrated in the somatic region and were absent from dendrites (Fig. 1). On the contrary, the mRNA for microtubule-associated protein 2, an established marker for dendritic mRNA, accumulated in distal dendrites (Fig. 1, arrowheads). This evidence agrees with previous studies (13,30) and suggests that GABA B R subunits are synthesized in the cell body and transported to dendrites via the secretory pathway.
We visualized the mobility of the subunits in live hippocampal neurons by FRAP using GABA B R1 and GABA B R2 fused to EGFP. Importantly, both EGFP-fused subunits maintained their trafficking properties relative to native subunits and efficiently coupled to G-protein inwardly rectifying potassium channels (supplemental Fig. S1 and data not shown). In dendrites GABA B R1-EGFP did not reach the plasma membrane Rat hippocampal neurons grown in culture were fixed at 14 DIV and processed for FISH with digoxigenin-labeled antisense probes to microtubule-associated protein 2 or the GABA B R subunits. The probes were detected using digoxigenin and secondary antibodies conjugated to Cy3. The outline of each neuron is drawn in white. Lower panels correspond to high magnification images of boxed areas above. and displayed an irregular fine punctate pattern ( Fig. 2A, left), consistent with its localization in the ER or ER-Golgi intermediate compartment in the absence of GABA B R2 (11,21,31). In contrast, the distribution of GABA B R2-EGFP was more heterogeneous, with larger tubulovesicular structures ( Fig. 2A,  right). GABA B R1-EGFP and GABA B R2-EGFP recovered efficiently after photobleaching revealing abundant pools of mobile subunits (Fig. 2A). However, the extent of recovery was different for GABA B R1 and GABA B R2. Although GABA B R1 recovered ϳ70%, GABA B R2 only reached ϳ50% of the original fluorescence (Fig. 2B).
We also performed time-lapse microscopy to explore the mobility of GABA B R subunits using mRFP fused to GABA B R1 and GABA B R2-EGFP (supplemental Fig. S1). Over a period of 30 min the majority of GABA B R1-mRFP-and GABA B R2-EGFP-containing structures moved independently and in bidi-rectional fashion (supplemental Fig. S2). These results indicate that GABA B R1 and GABA B R2 are transported in dendrites and suggest that they reside in segregated compartments, which move with different kinetics.
Segregated GABA B R1 and GABA B R2 Subunits Are Abundant in the Brain-To determine whether a significant pool of nonassembled GABA B R subunits exist in intracellular compartments of the brain we complemented our analysis with a biochemical approach. We coimmunoprecipitated GABA B R1 and GABA B R2 from two subcellular rat brain fractions enriched in different membranous organelles (P2, a fraction enriched in plasma membrane and P3, a fraction enriched in light intracellular membranes). Calnexin, an ER resident protein, was enriched in P3 confirming the effectiveness of the differential centrifugation procedure (Fig. 3B). Although the net amount of GABA B R1 that coimmunoprecipitated with GABA B R2 was similar in P2 and P3, the total subunit abundance was significantly lower in P2 than in P3 (Fig. 3, A-D). Thus, the relative abundance of associated subunits was significantly higher in P2 than in P3 (Fig. 3E). These observations demonstrate that the proportion of non-associated subunits is higher in intracellular compartments than in the plasma membrane in the brain.
Segregated GABA B R1 and GABA B R2 Subunits Predominate in Dendritic Intracellular Compartments-We then used quantitative colocalization to discriminate between associated and segregated GABA B R1 and GABA B R2 subunits in dendrites of cultured hippocampal neurons. GABA B R subunits were distributed as densely packed granules, with minor differences during neuronal differentiation (Fig. 4A). Merged confocal images revealed a high proportion of non-colocalized subunits (Fig. 4A). A quantitative analysis yielded low colocalization between GABA B R1 and GABA B R2 for all stages of differentiation (Fig. 4 These results demonstrate that a high proportion of GABA B R1 and GABA B R2 are segregated in dendrites. Moreover, taking into account that hippocampal neurons actively establish synapses during the first weeks in culture (not shown), they demonstrate that the proportion of subunit association is independent of synaptogenesis. Similar results were obtained in axons (supplemental Fig. S3).
Colocalization was high when a control protein, the GluR2 subunit of the AMPA receptor, was detected with a single primary antibody and two different secondary antibodies, validating the quantitative colocalization method in dendrites (Fig. 4, B-D, M 1 GluR2 green/red: 0.788 Ϯ 0.018; M 2 GluR2 red/green: 0.785 Ϯ 0.031). In addition, high and almost identical coefficients were obtained when comparing the colocalization of GABA B R1 and GABA B R2 at the plasma membrane of COS-7 cells using antibodies to endogenous C-terminal epitopes or extracellular N-terminal epitope tags (not shown). Furthermore, no antibody-induced clustering was detected (not shown). These results indicate that the antibodies are specific and appropriate for a colocalization analysis and that the analysis is not affected by epitope masking. To provide an independent measurement of the proportion of segregated subunits we used recombinant Myc-GABA B R1 and HA-GABA B R2 in hippocampal neurons. Importantly, the distribution patterns of recombinant and endogenous subunits were indistinguishable (Fig. 4, A and E). By detecting the N-terminal epitope tags under non-permeabilized conditions we had access exclusively to the subunits at the cell surface. Alternatively, by detecting the same epitopes under permeabilized conditions we evaluated the total population of cell surface plus intracellular subunits. As expected, GABA B R1 and GABA B R2 displayed high colocalization at the dendritic plasma membrane (Fig. 4, E (upper panel) and G, M 1 GABA B R1/R2: 0.609 Ϯ 0.014, M 2 GABA B R2/R1: 0.645 Ϯ 0.015, n ϭ 5 cells, 5 slices each). In contrast, the colocalization was significantly lower when the intracellular populations were included in the analysis (Fig. 4, E (lower panel) and G, M 1 GABA B R1/R2: 0.355 Ϯ 0.029, M 2 GABA B R2/R1: 0.378 Ϯ 0.034, n ϭ 5 cells, 5 slices each). We also analyzed the colocalization between GABA B R1 and a chimera containing the N terminus and seven transmembrane domains of GABA B R2 fused to the C terminus of GABA B R1 (GABA B R2 R1C ), including the ER retention motif (22). Colocalization between this pair was significantly higher than control (Fig. 4, F and G).
To support these observations we evaluated the intracellular colocalization of endogenous GABA B R subunits. We could not directly monitor the surface/intracellular ratio for native GAB-A B Rs due to the lack of antibodies against extracellular epitopes. However, we measured the differences between a region "near the plasma membrane" (PM) and the "core region" (CR) of a dendrite. We discriminated between PM and CR through image processing by isolating 0.6-m strips along the margins of the dendrite. Colocalization was significantly higher for PM than CR (Fig. 4H, M 1  A visual examination also revealed that GABA B R subunits showed little colocalization in the cell body (Fig. 4I). Although GAB-A B R1 was prominent in a ring around the nucleus and displayed a homogeneous pattern in the rest of the soma, GABA B R2 was heterogeneous with abundant staining near the cell surface and in the perinuclear area. Taken together these observations reveal the prevalence of segregated subunits in intracellular compartments of cell bodies and dendrites. They also indicate that the C termini of GABA B R1 and GABA B R2 are essential to keep the subunits segregated in intracellular compartments of dendrites.

Heteromeric Receptors Are Preferentially Associated to the Plasma Membrane, but Disruption of ER Export Results in Intracellular Somatodendritic
Accumulation-We took advantage of BiFC to unambiguously localize the assembled heteromer within the neuron. GABA B R1 was fused to YFP-N, whereas GABA B R2 was fused to YFP-C. GABA B R-dependent BiFC signal (GABA B R-BiFC) was only observed when neurons were cotransfected with Myc-GAB-A B R1-YFP-N and HA-GABA B R2-YFP-C, and not when the constructs were expressed independently (supplemental Fig.  S4). GABA B R-BiFC occurred predominantly at the plasma membrane of somatic and dendritic compartments as indicated by the overlap with simultaneous surface staining of GABA B R1 (Fig. 5A). Occasionally GABA B R-BiFC accumulated in intracellular compartments where it did not colocalize with the ER or the Golgi apparatus (Fig. 5, B and C). These observations are in agreement with the lack of subunit colocalization in intracellular compartments of dendrites.
We then used BFA to block the ER to Golgi transport (32). Acute BFA treatment caused the dispersal of the Golgi apparatus but had no effect on GABA B R-BiFC localization (Fig. 6, A  and B). This suggests that the pool of assembled GABA B Rs in transit through the secretory pathway is small compared with the levels of heteromers at plasma membrane at steady state. To capture this limited in-transit pool we incubated neurons for a prolonged period with a low concentration of BFA (33). As expected, the Golgi apparatus was dispersed after BFA treatment, but now the majority of GABA B R-BiFC remained within intracellular compartments in the cell body and dendrites (Fig. 6C).  4) were prepared from adult rat brains. Samples were immunoprecipitated with control IgG (lanes 1 and 3) or GABA B R2 antibodies (lanes 2 and 4), separated by SDS-PAGE and immunoblotted with GABA B R1 antibodies. B, the P2 and P3 fractions prior to immunoprecipitations were used to control the abundance of GABA B R1, GABA B R2, and calnexin. C, immunoblots for immunoprecipitations of GABA B R1 from P2 or P3 were analyzed by densitometry and average values Ϯ S.E. were plotted for each fraction (n ϭ 3 independent fractionation and immunoprecipitation experiments). D, same as above for the total abundance GABA B R1 in each fraction. E, values of immunoprecipitations were normalized to the abundance of GABA B R1 in the corresponding fraction and average values Ϯ S.E. were plotted for each fraction (***, p Ͻ 0.001).
Finally, we expressed GABA B R-BiFC in the presence of Sar1[H79G], a dominant negative form of Sar-1 that is locked in a GTP-bound state and blocks export from the ER (32). Impor-tantly, in the presence of Sar1[H79G] GABA B R-BiFC accumulated intracellularly in the soma and dendrites (Fig. 6D). Consistently, the colocalization between GABA B R-BiFC and the ER was significantly increased in the presence of Sar1[H79G] indicating that GABA B R-BiFC was retained in the ER in dendrites (Fig. 6, E and F). Taken together these observations strongly support the notion that GABA B R subunits assemble in the ER throughout the somatodendritic compartment.

DISCUSSION
The existence of heteromeric GABA B Rs at the cell surface of neurons has been convincingly demonstrated by biochemical, functional, and microscopic observations (16, 21, 34 -37). Likewise, the ER retention of GABA B R1 and the necessary assembly with GABA B R2 have been shown in a variety of cellular contexts (9). However, the relationship between GABA B R trafficking and the geometrical arrangement of the secretory pathway in neurons has received limited attention. Our findings conclusively show that GABA B R subunits are abundant and segregated in intracellular compartments. They provide compelling evidence for the absence of long haul post-Golgi vesicles containing assembled GABA B Rs in dendrites. They do not challenge the functional heteromerization hypothesis. Rather, they suggest that assembly of GABA B R1 and GABA B R2 and export from the ER are processes that occur throughout the somatodendritic compartment.
Combined with previous reports a model emerges to explain the availability of GABA B Rs at the plasma membrane in neurons. First GABA B R subunits are synthesized in the neuronal cell body and move along dendritic ER membranes. Because GABA B R2 progresses freely through the secretory pathway (9) it picks up ER-retained GABA B R1, and the resulting heteromer rapidly exits the ER in the cell body and dendrites. Continuously, the assembly of intracellular subunits contributes to the turnover of the surface receptors. Once at the plasma membrane GABA B Rs are endocytosed as heteromers in a constitutive manner and undergo efficient recycling (38).
The Presence of Segregated Subunits Defines the Mode of Receptor Trafficking-Our results suggest the existence of negative regulatory factors such as chaperones or specialized subdomains of the ER that prevent heteromeric assembly immediately after synthesis in the cell body. The differential distribution of resident proteins in sub-domains of the ER has been reported before. For example calsequestrin, a luminal resident of the ER, distributes differentially depending on its oligomeric state (39). Regarding ER regulatory proteins 14-3-3 and coatomer complex protein I have been proposed to participate in the ER retention of GABA B Rs (31). Several other proteins interact with the receptor subunits throughout the secretory pathway (e.g. msec7-1, C/EBP homologous protein, Marlin-1, activating transcription factor 4, and G protein-coupled receptor kinase 4) and may block their association temporarily (40). In addition, the ER retention and trafficking properties of GABA B receptors are modulated by the distinct class of GABA A receptors, raising the possibility of physical and functional cross-talk between these ionotropic and metabotropic neurotransmitter receptor systems (41). However, it remains to be established how ER segregation or GABA B R-associated proteins affect trafficking or whether they contribute to regulate the assembly of GABA B Rs. It is also not clear whether recently FIGURE 4. GABA B R1 and GABA B R2 are segregated in intracellular compartments of primary dendrites. A, hippocampal neurons were labeled with antibodies to GABA B R1 (green) and GABA B R2 (red). Images were merged to visualize colocalization in dendrites (right). Drawings on the right: circles correspond to native epitopes; black, GABA B R1; empty, GABA B R2; red and green symbols correspond to secondary antibodies. B, same as above in neurons labeled with primary antibodies to GluR2 and two secondary antibodies. Images were merged to visualize colocalization (right). Drawings on the right: circles correspond to native epitopes; gray, GluR2; red and green symbols correspond to secondary antibodies.  assembled heteromers follow a conventional or a local secretory route after ER export and how ER exit is regulated (42).
Most of the evidence regarding GABA B Rs indicates that the subunits assemble in the ER (9). However, recent reports suggest that GABA B R1 reaches the cis-Golgi or the trans-Golgi network (31,36). These observations imply that assembly of the heteromer occurs not in the ER but in a post-ER membrane organelle. Our results are in disagreement with these findings, because upon blockade of ER export assembled GABA B Rs accumulate throughout the somatodendritic ER.
Although electron microscopy studies indicate that GAB-A B R1 and GABA B R2 localize to similar intracellular membranes in the visual cortex, the degree of intracellular colocalization is lower than at pre-or postsynaptic membranes, supporting our conclusions (43). The differential accumulation of GABA B R subunits in intracellular compartments is not unique and may be common for a variety of multisubunit receptors requiring tight assembly and precise targeting control. For example, the GluR1 and GluR2 subunits of the AMPA receptor accumulate abundantly in intracellular compartments and display significant differences in speed and direction of intracellular mobility (44). In addition, AMPA receptor subunits distribute differentially in intracellular compartments and their distribution coincides with tubulovesicular membranes in the dendritic ER (45). Interestingly, N-methyl-D-aspartic acid receptors are also exported from the ER in dendrites (46). Thus, a conserved mechanism may exist to regulate the transport and assembly of receptors through the ER membrane and their assembly en route to the synapse.
Implications for Receptor Availability at the Plasma Membrane-GABA B R-induced currents increase significantly between 5 and 14 DIV in cultured hippocampal neurons (34). Unexpectedly our results demonstrate that the proportion of colocalized GABA B R subunits, an indirect indication of functional receptors, does not change during differentiation. These observations are surprising considering the robust synaptogenic conditions of the culture. Two alternative interpretations of these results are possible. First, the number of associated subunits may increase proportionally to the number of segregated ones. Second, the abundance of membrane receptors may be constant during the initial stages of synaptogenesis, but the gradual acquisition of functional GABA B Rs may be defined by the availability of downstream signaling molecules. Taking into account the continuous increase in the abundance of total GABA B R1 and GABA B R2 during the first 21 DIV, the increase in surface GABA B R2 during the same period, and the constant levels of G-protein inwardly rectifying potassium channel 1 (34), an established downstream signaling molecule for GABA B Rs, we favor the first alternative and suggest a continual and proportional replenishing of associated GABA B Rs from intracellular segregated pools.
GABA B Rs are also abundant in axons, but their mode of transport remains unexplored (16,17). Our colocalization analysis also suggests the more controversial mechanism that segregated subunits assemble in axons. This possibility poses additional challenges: although the existence of the ER in axons has been reported (47), the presence of late secretory components is less clear. Therefore, understanding the precise assembly mechanism of GABA B Rs and other multisubunit receptors in axons requires a detailed characterization of the complex morpho-topological distribution of secretory organelles in neurons (48).