Endoplasmic Reticulum Sorting and Kinesin-1 Command the Targeting of Axonal GABAB Receptors

In neuronal cells the intracellular trafficking machinery controls the availability of neurotransmitter receptors at the plasma membrane, which is a critical determinant of synaptic strength. Metabotropic γ amino-butyric acid (GABA) type B receptors (GABABRs) are neurotransmitter receptors that modulate synaptic transmission by mediating the slow and prolonged responses to GABA. GABABRs are obligatory heteromers constituted by two subunits, GABABR1 and GABABR2. GABABR1a and GABABR1b are the most abundant subunit variants. GABABR1b is located in the somatodendritic domain whereas GABABR1a is additionally targeted to the axon. Sushi domains located at the N-terminus of GABABR1a constitute the only difference between both variants and are necessary and sufficient for axonal targeting. The precise targeting machinery and the organelles involved in sorting and transport have not been described. Here we demonstrate that GABABRs require the Golgi apparatus for plasma membrane delivery but that axonal sorting and targeting of GABABR1a operate in a pre-Golgi compartment. In the axon GABABR1a subunits are enriched in the endoplasmic reticulum (ER), and their dynamic behavior and colocalization with other secretory organelles like the ER-to-Golgi intermediate compartment (ERGIC) suggest that they employ a local secretory route. The transport of axonal GABABR1a is microtubule-dependent and kinesin-1, a molecular motor of the kinesin family, determines axonal localization. Considering that progression of GABABRs through the secretory pathway is regulated by an ER retention motif our data contribute to understand the role of the axonal ER in non-canonical sorting and targeting of neurotransmitter receptors.


Introduction
Polarized protein trafficking in the neuron is critical for synapse formation, synapse maintenance and the regulation of synaptic strength. In all eukaryotic cells the endomembrane trafficking system includes a forward biosynthetic route constituted by the endoplasmic reticulum (ER), the ER-Golgi intermediate compartment (ERGIC), the Golgi apparatus and post-Golgi vesicles, and a recycling-degradative route constituted by endosomes and lysosomes. The unique architecture and size of neurons does not necessarily imply that the structure/function relationship of these organelles and their contribution to the secretory process are different than in other cell types. However, their spatial arrangement and contribution to local processing may be specially adapted to the complexities of the neuronal morphology [1]. How the neuron orchestrates this highly compartmentalized trafficking is poorly understood. In particular, how the local distribution of secretory components in the neuron impinges on intracellular trafficking and availability of neurotransmitter receptors remains for the most part unexplored.
GABA is the main inhibitory neurotransmitter in the nervous system and the metabotropic GABA B Rs are obligatory heteromers composed of two related subunits, GABA B R1 and GABA B R2 (for a comprehensive review of GABA B R structure, function, localization and pathological implications see [2]). Both belong to family C of G protein-coupled receptors, and contain a large extracellular N-terminal domain, seven membrane-spanning domains and an intracellular C-terminal domain. GABA B Rs are expressed in neurons throughout the brain and spinal cord. They are mainly perisynaptic receptors located in gabaergic and glutamatergic presynaptic terminals and postsynaptic sites. GABA B R1 binds agonists with high affinity whereas GABA B R2 couples to G ai establishing a transactivation mechanism between the two subunits. At presynaptic terminals GABA B Rs inhibit voltage gated Ca 2+ channels thereby inhibiting synaptic vesicle fusion and neurotransmitter release. At postsynaptic sites they activate inwardly rectifying K + channels hyperpolarizing the postsynaptic neuron. In addition, stimulation of GABA B Rs decreases the levels of cyclic AMP. GABA B Rs have been implicated in epilepsy, anxiety, stress, sleep disorders, nociception, depression, cognition and addictive mechanisms to drug abuse. The relevance of studying GABA B R availability is further supported by clinical observations that report the appearance of tolerance to GABA B R agonists, an inconvenient side effect to therapy.
GABA B R subunits are synthesized in the soma and glycosylated in the ER [3], [4]. The progression of GABA B Rs through the secretory pathway is regulated by an RXR-type sequence (RSRR) in the C-terminal domain of GABA B R1 that functions as an ER retention motif in the absence of GABA B R2 [5]. The ER retention motif is masked upon association to GABA B R2, and assembled GABA B Rs exit the ER as heteromers destined for the plasma membrane. Consistent with ER retention acting as a limiting step GABA B Rs are abundant within intracellular compartments, especially the ER [6].
GABA B R1a and GABA B R1b constitute the most abundant isoforms for GABA B R1. Heteromers containing GABA B R1a are axonal and somatodendritic whereas those containing GA-BA B R1b are exclusively located in the somatodendritic domain [7]. GABA B R1a and GABA B R1b mediate their different functions only as a result of their specific axonal or somatodendritic localization [7]. The sushi domains located at the Nterminus of GABA B R1a are necessary and sufficient for axonal targeting even in a GABA B R2 knock-out background [8].
However, the precise targeting machinery and the organelles involved in sorting and transport have not been described. Combining conventional optical microscopy and live-cell imaging using organelle reporters and trafficking blockers in cultured hippocampal neurons we describe a mechanism for GABA B R1a axonal localization based on pre-Golgi sorting and ER transport.

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 (The National Academy of Sciences, 1996). GABA B R1-EGFP mice were kindly provided by Bernhard Bettler (University of Basel, Switzerland). They correspond to transgenic animals for the GABA B R1-EGFP BAC in a homozygous knockout background for GABA B R1 as described previously [9], [10].
Cell lines, neuronal cultures and transfection COS-7 cells were maintained and transfected as described previously [3] using a Gene Pulser Xcell (BioRad). Primary hippocampal neurons were cultured from E18 rats or E18 GABA B R1-EGFP transgenic mice according to established procedures [11] and transfected by Ca 2+ phosphate at 14-18 days in vitro (div) [12]. All transfected neurons, except for those in Fig. S2, were analyzed 1 day post transfection (dpt). Transfected neurons for Fig. S2 were analyzed between 1-5 dpt.

Reagents and DNA plasmids
Nocodazole was purchased from Sigma (St. Louis, MO). MYC-GABA B R1, FLAG-GABA B R2, HA-GABA B R2, MYC-GA-BA B R1-AA-ASA, MYC-GABA B R1-DC in pRK5 have been described previously and contain epitope tags on the extracellular N-terminal domains [3], [13][14][15]. MYC-GABA B R1-AA-ASA contains point mutations that replace two leucine residues at a di-leucine motif and two arginine residues at the ER retention motif (RSRR) by alanines. MYC-GABA B R1-DC lacks the complete C-terminal domain. Both mutants escape the ER and traffic to the cell surface in the absence of GABA B R2. GABA B R1a-EGFP, GABA B R2-EGFP and GABA B R1a-monomeric red fluorescent protein have also been described previously and contain the fluorescent proteins attached to the intracellular C-terminal domain [4]. pDsRed-C1 (RFP), pEYFP-Golgi, pEYFP-ER and pDsRed2-ER (KDEL-RFP) were obtained from Clontech (Mountain View, CA). Kif5C-RFP-DN was kindly provided by S. Kindler and H.J. Kreienkamp (Institut für Humangenetik, Universitä tsklinikum Hamburg-Eppendorf, Hamburg, Germany) and corresponds to amino acids 678-955 of KIF5C (NM_001107730) also referred to as DN2 by Falley and collaborators [16]. ARF1-Q71I-HA was kindly provided by O. Jeyifous (University of Chicago, Chicago, IL), Rab11-GFP was kindly provided by F. Bronfman (Pontificia Universidad Católica de Chile), p58-YFP was kindly provided by J. Lippincott-Schwartz (National Institutes of Health, Bethesda, MD). Cytochrome b5-EGFP was kindly provided by C. Hetz (Universidad de Chile). All manipulations and fidelity of DNA constructs were verified by sequencing.
Antibodies GABA B R1 antibodies (which recognize GABA B R1a and GABA B R1b) were raised against the intracellular C-terminal domain in rabbits and affinity purified. GABA B R1 antibodies specifically recognize MYC-GABA B R1a in transfected COS cells and detect the predicted doublet corresponding to GABA B R1a and GABA B R1b in crude brain membranes (Fig. S1). Microtubule-associated protein 2 (MAP2) antibodies were purchased from Chemicon (Temecula, CA). Piccolo antibodies were kindly provided by ED. Gundelfinger and WD. Altrock (Leibniz Institute for Neurobiology, Magdeburg Germany). KDEL antibodies (directed against a 6-residue peptide (SEKDEL) of the rat Grp78 protein) and cis-Golgi matrix protein 130 (GM130) were purchased from StressGen (Ann Arbor, MI). MYC antibodies were purchased from Sigma (St. Louis, Missouri). Influenza A Virus epitope (HA) antibodies were purchased from Roche (Indianapolis, IN). Anti-GFP antibodies (ab6556) were purchased from Abcam (Cambridge, UK). Secondary anti-mouse, antirabbit, anti-guinea pig or anti-chicken antibodies conjugated to Texas Red (TR), tetramethyl rhodamine isothiocyanate (TRITC), fluorescein isothiocyanate (FITC) or cyanine 3 (Cy3) were purchased from Jackson Immuno Research Laboratories (West Grove, PA).
Immunofluorescence, image capture, image processing and time-lapse microscopy Immunofluorescence was performed as described previously under non-permeabilized or permeabilized conditions [4]. Depending on the type of experiment axons were identified by the presence of a positive marker (Tau), the absence of a negative marker (MAP2), or by morphological criteria that included: longer projection, constant diameter and right angle branching. Imageprocessing routines were developed in our laboratory based on of Interactive Data Language (IDL) (ITT, Boulder, CO), including routines for segmentation [4]. Live-cell imaging was performed in a 23uC equilibrated microscopy suite. Images were obtained using an Olympus BX61WI upright microscope and an Olympus diskscanning unit. Consecutive frames were acquired over a period of ,120 s. Kymographs were constructed from an axonal segment 20 mm from the soma using ImageJ from three-pixel-wide axonal traces. The axon-to-dendrite (A:D) ratio of MYC-GABA B R1a was determined using ImageJ according to a procedure based on Biermann et. al., 2010 and modified [8]. Briefly, one-pixel-wide lines were traced along the initial 150 mm of axons and dendrites in images labeled with soluble RFP or Kif5C-RFP-DN. No measurements were carried out beyond 150 mm to prevent artifacts due to axonal length differences between RFP or Kif5C-RFP-DN transfected neurons. Average pixel intensities of MYC-GABA B R1a were determined along the traced lines, background was subtracted and the data was used to determine the A:D ratio. Criteria for cell selection included even distribution of RFP or Kif5C-RFP-DN in axons and dendrites, and cells expressing constructs at very high levels were excluded from the analysis. 9-12 neurons from at least two independent culture preparations were analyzed for each condition. Images of axons from GABA B R1-EGFP mice were acquired using an Olympus FV-1000 confocal microscope. Anti-GFP antibodies were used to amplify the weak axonal signal of GABA B R1-EGFP expressed at physiological levels.

Total Internal Reflection Fluorescence (TIRF) microscopy
For TIRF microscopy 14-18 div hippocampal neurons were transfected with GABA B R1a-EGFP and motility in axons was analyzed 1 dpt. TIRF was carried out on a custom-built TIRF microscope setup: Two lasers (473 nm and 532 nm, both 30 mW, Viasho, USA) were used to excite the fluorophores (GFP and RFP). The lasers were expanded and coupled via a multi-beam splitter (z474/488/532/635rpc, Chroma, USA) off-axis into the oil-immersion objective (Nikon, SFluor 100x, 1.49) to obtain TIRF illumination. The emitted fluorescent light was split in the GFP and RFP signals using a dichroic mirror (525/50, Chroma, USA) then passed trough bandpass filters (530/50 for GFP and 605/70 for TMR, both Chroma, USA) and finally directed via mirrors to separate areas on the chip of a frame transfer CCD camera (Cascasde:512B, Roper Scientific, USA). The CCD camera was controlled via WinSpec32 (Princeton Instruments, USA). The penetration depth of TIRF was 147 nm. Digital images were taken at a frame rate of 2 frames/s and were subsequently analyzed for velocity and direction using kymographs generated with a customwritten LabView (National Instruments, USA) routine. Kymographs were analyzed for velocity and direction by fitting lines to the segments of a trace judged by eye. Stalls were not taken into account.

Results
The delivery of GABA B Rs to the plasma membrane is Golgi-dependent but axonal targeting is not First we carried out a control experiment to validate the use of overexpression of recombinant GABA B R subunits as a strategy to study receptor trafficking. Cultured hippocampal neurons were transfected with MYC-GABA B R1a and the distribution of the subunit at the plasma membrane or in intracellular compartments was evaluated by immunostaining 1-5 dpt. MYC-GABA B R1a was retained in intracellular compartments in the cell body and axons up to 5 dpt in the absence of recombinant GABA B R2 expression (Fig. S2). In contrast, GABA B R1a was readily detectable at the cell surface at 2 dpt upon co-transfection with GABA B R2 (Fig. S2, right column). These experiments indicate that the trafficking properties of recombinant receptors mimic the situation of the native subunits, and that the trafficking of recombinant receptors is not affected by the endogenous subunits. More importantly, they demonstrate that our experiments using transfection of recombinant GABA B R1 subunits exclusively examine their intracellular population.
To determine whether GABA B Rs employ a Golgi-dependent intracellular trafficking route in neurons, primary cultures of hippocampal neurons were transfected with MYC-GABA B R1a and FLAG-GABA B R2 in the absence or presence of ARF1-Q71I-HA, a constitutively active ARF1 mutant that prevents export from the Golgi apparatus [17]. 1 dpt the distribution of GABA B Rs at the plasma membrane and in intracellular compartments was evaluated by immunostaining under nonpermeabilized or permeabilized conditions. We examined somatic or axonal domains as shown in the schematic neuron (Fig. 1A). As reported previously co-transfection of MYC-   (Fig. 1B). In contrast, ARF1-Q71I-HA blocked the appearance of both subunits at the plasma membrane and produced accumulation in intracellular compartments (Fig. 1C). These results indicate that the Golgi apparatus is necessary for the delivery of GABA B Rs to the plasma membrane in hippocampal neurons.
GABA B R1a is targeted to the axon in hippocampal neurons [7], [8]. Thus, we determined whether axonal targeting was also Golgi dependent. First we used cultured hippocampal neurons of velocities: anterograde (gray bars), retrograde (white bars). Pie chart represents fractions of anterograde (gray) and retrograde transport (white). Average velocities and direction were obtained from 41 moving puncta from at least three independent culture preparations. doi:10.1371/journal.pone.0044168.g003  1D and 1G, right panels). According to previous reports the predominant axonal variant of GABA B R1 corresponds to GABA B R1a [7], [8]. Therefore, these data are compatible with the idea that GABA B R1a is sorted and targeted to the axon at or prior to the Golgi apparatus. Next we used recombinant receptors to directly compare the axonal targeting of GABA B R1a and GABA B R1b, and their Golgi dependence. Neurons were transfected with MYC-GABA B R1a or MYC-GABA B R1b in the absence or presence of ARF1-Q71I-HA. Consistent with previous reports, GABA B R1a but not GABA B R1b was predominantly targeted to the axon in hippocampal neurons (Figs. 1E-1I, control panels). Importantly, recombinant GABA-B R1a was still targeted to the axon in the presence of ARF1-Q71I-HA (Figs. 1E and 1H). In contrast, GABA B R1b was absent from the axon under all the conditions examined (Figs. 1F and 1I). These findings indicate that axonal targeting is specific to GABA B R1a. In addition, they demonstrate that the sorting and targeting of GABA B R1a to the axon occurs at or prior to the Golgi stage, and therefore suggest a non-conventional modality. They also indicate that axonal targeting of GABA B R1a upon ARF1-Q71I-HA expression is not produced by an overload of the early secretory pathway because the effect was not observed for GABA B R1b, a conclusion further supported by the results in lower expressing transgenic neurons.

GABA B R1a is targeted and transported within the axonal ER
ER resident proteins and components of the protein folding and export machineries localize to the axon [18][19][20]. Thus, to determine the intracellular localization of axonal GABA B R1a neurons were immunostained with antibodies against the GA-BA B R1 subunit and an antibody against the SEKDEL sequence of the rat ER protein Grp78, including a conserved motif present in luminal ER resident proteins responsible for retrieval from the Golgi apparatus [21]. Endogenous GABA B R1 colocalized with the ER in axons ( Fig. 2A, arrows). Control labeling without primary antibodies confirmed the specificity of the signal (Fig. 2B and 2C). In transgenic mouse neurons GABA B R1-EGFP also colocalized with the ER (Fig. 2D, arrows). Importantly, colocalization was still observed upon overexpression of MYC-GABA B R1a and KDEL-RFP, a fluorescent probe widely used for ER visualization that contains the ER targeting sequence of calreticulin and the ER retrieval sequence KDEL (Fig. 2E, arrows). In addition, colocalization was observed with cytochrome b5-EGFP, another fluorescent ER probe (Fig. 2F). The colocalization between GABA B R1a and ER markers was specific because Piccolo, a marker for dense core vesicles and synapses [22] showed a markedly different axonal localization (Fig. 2G). Staining with MAP2, an exclusive dendritic marker confirmed the co-distribution of GABA B R1a and the ER occurs in the axon (Fig. 2H). These results demonstrate that GABA B R1a is enriched in the axonal ER.
To establish whether the ER functions as a transport organelle for GABA B R1a we examined the dynamic behavior of fluorescent versions of GABA B R1a and the ER in axons of live hippocampal neurons. Discrete GABA B R1a-EGFP and KDEL-RFP puncta were distributed along the axons. While the majority of puncta remained static or showed very limited lateral displacement over the examined period, a subset displayed continuous, long-range mobility (imaged at 0.20-0.25 frames/s for a total of ,120 s, Figs. 3A and 3B). GABA B R1a-EGFP and KDEL-RFP moved bidirectionally, with a moderate retrograde bias, and with modal speeds of 100-200 nm/s (Figs. 3E and 3F). Significantly, some GABA B R1a-EGFP and KDEL-RFP puncta moved in synchrony, with similar speed and retrograde predominance (Figs. 3C and 3G). Puncta containing Rab11-GFP, a recycling endosome marker [23], also localized to axons but showed a different dynamic pattern characterized by rapid direction changes and lower overall displacement. We used TIRF microscopy and higher temporal resolution (2 frames/s) to determine the instant velocity of GABA B R1a-EGFP more accurately in hippocampal neurons. Mean anterograde and retrograde instant velocities were comparable (751.70633.20 and 877.73667.59 nm/s respectively) (Figs. 3H and 3I). These values fit conventional kinesin velocities [24] and their slight increase above panels A-G most likely result from excluding stalls in the analysis of our higher temporal resolution imaging.
A proportion of GABA B R1a-EGFP and KDEL-RFP puncta moved independently from each other. This may indicate that a fraction of GABA B R1a-EGFP is transported in a different secretory organelle or that the axonal ER compartment is capable of dynamically segregating cargo. To discriminate between these possibilities we carried out a series of complementary experiments. First we analyzed time-lapse microscopy sequences individually. Interestingly, GABA B R1a-EGFP puncta that initially colocalized with the ER sometimes separated from the organelle, remained segregated for a few frames and fused again with a pre-existing ER compartment (Fig. 4A). It is well known that ER cargo recycles between the ER and the ERGIC using export/retrieval motifs such as the RXR-type sequence present in GABA B R1a [25]. To determine whether the segregated GABA B R1a puncta resided temporarily in the ERGIC, we first visualized the axonal distribution of MYC-GABA B R1a and p58-YFP, an established marker of the ERGIC [26], in fixed cells. Sparse ERGIC puncta were observed in axons and a subset of them colocalized with MYC-GABA B R1a (Fig. 4B). Additionally, live-cell imaging was carried out in neurons transfected with GABA B R1a-RFP and p58-YFP. A small fraction of GABA B R1a-RFP displayed synchronous motility with p58-YFP in axons (Fig. 4C). Taking into account that intra-ER mobility is microtubule dependent but short-range ER to ERGIC transport is not [27], we reasoned that the enrichment of GABA B R1a in the ER should increase upon destabilization of microtubules. Consistent with this prediction, nocodazole blocked the mobility of GABA B R1a-EGFP puncta and the subunits accumulated in a KDEL-RFP compartment (Fig. 4D, top right panels). As expected, a mutant GABA B R1a subunit that is not retained in the ER (GABA B R1a-ASA-EGFP) accumulated in a different compartment after nocodazole treatment (Fig. 4D, bottom right panels). These observations suggest that axonal GABA B R1a is targeted and transported to the axon within the ER, and possibly engages in a local export/retrieval mechanism between the ER and the ERGIC.

Kinesin-1 contributes to the axonal localization of GABA B R1a
Since kinesin-1, an axonal biased molecular motor [28], [29], colocalizes with GABA B R1 in neurons, and associates to the subunit in fractionation and coimmunoprecipitation assays [30] we evaluated its contribution to axonal GABA B R1a targeting. To examine the axonal localization of MYC-GABA B R1a we determined an axon-to-dendrite ratio (A:D ratio) [8]. To study the role of kinesin-1 we used a dominant negative comprising the cargo-binding domain of Kif5C that interferes with endogenous kinesin-cargo interactions fused to RFP (Kif5C-RFP-DN) [16]. The axonal targeting of MYC-GABA B R1a was markedly reduced in the presence of Kif5C-RFP-DN ( Fig. 5A and Fig. S3 Kinesin-1 may control the axonal localization of MYC-GABA B R1a via an adaptor mechanism through the cytosolic C-terminal domain of GABA B R1a. To directly test this we first determined whether axonal targeting required the cytosolic Cterminal domain of GABA B R1a using two C-terminal subunit mutants, MYC-GABA B R1a-AA-ASA and MYC-GABA B R1a-DC [15]. GABA B R1a, GABA B R1a-AA-ASA and GABA B R1a-DC were all abundant in the axon indicating that targeting is independent of the C-terminal domain (Fig. 5C, arrows). Using a quantitative colocalization analysis based on Manders coefficients [31] all GABA B R1a constructs colocalized partially with the ER and no statistically significant difference was observed between GABA B R1a and the two mutants (MYC-GABA B R1a, n = 9 neurons; MYC-GABA B R1a-AA-ASA p = 0.11, n = 5 neurons; MYC-GABA B R1a-DC p = 0.30, n = 5 neurons). These results suggest that the C-terminal domain is not a major determinant of axonal ER distribution and targeting of GABA B R1a.
Finally, we analyzed the kinesin-1 dependence on the axonal targeting of GABA B R1a lacking the C-terminal domain. Axonal localization of MYC-GABA B R1a-DC was still severely impaired by the dominant negative construct (Fig. 5D, arrows. A:D ratio MYC-GABA B R1a-DC 0.4260.06; MYC-GABA B R1a-DC plus Kif5C-RFP-DN 0.0560.01; p,0.01). Combined these experiments demonstrate that kinesin-1 is necessary for the ER axonal targeting and localization of GABA B R1a. In addition, they indicate that the C-terminal domain of GABA B R1a, which is exposed to the cytosol, is not involved in this transport mechanism, further supporting the role of luminal or ER membrane domains in axonal sorting and targeting.

Discussion
We have shown that GABA B Rs require the Golgi apparatus for plasma membrane delivery. Importantly, to our knowledge we have demonstrated for the first time that the sorting and targeting of an axonal neurotransmitter receptor, namely the GABA B R1a subunit, occur in a pre-Golgi compartment. Consistent with these observations, our evidence points to the fact that GABA B R1a traffics along the axonal ER and the ERGIC, and is transported by the molecular motor kinesin-1.

Pre-Golgi sorting and targeting of axonal GABA B R1a
Our study indicates that axonal sorting of GABA B R1 subunits operates in the ER. As reported elsewhere, axonal targeting of GABA B R1a is unaltered in GABA B R2 knock-out neurons [8]. Additionally, the sushi domains located at the N-terminus of GABA B R1a, are sufficient for targeting even when placed in a nonrelated CD8a protein context [8]. Combined with the results presented here these observations imply that sorting and targeting signals exposed to the ER lumen mediate the axonal localization of GABA B R1a.
Historically, sorting of plasma membrane proteins to distinct membrane domains has been thought to occur exclusively at the Golgi or the trans-Golgi network, but accumulated evidence now favors the view that decisions are made at almost every step along the secretory pathway including the ER [32]. For example, mutations in Sec24p, a component of the coat protein complex II (COPII), selectively disrupt recruitment of cargo for ER to Golgi transport [33]. Likewise, synthetic cell penetrating peptides based on the cytosolic domain of the temperature-sensitive VSVG protein only inhibits the transport of a subset of cargo from the ER to the Golgi apparatus [34].
Two possible targeting mechanisms are conceivable for GABA B R1a in axons: (i) a luminal or membrane spanning ER protein enriched in the axonal ER subcompartment may bind the sushi domains and produce the accumulation of GABA B R1a but not GABA B R1b, in the axon (selective retention); or (ii) a protein or protein complex that spans the ER membrane may function as an adaptor between the GABA B R1a subunit and a molecular motor and selectively direct the transport of the subunits to axons (selective transport). Additional axonal scaffolding proteins may anchor the GABA B R1a subunit to strengthen axonal localization.
Although we cannot rule out the first alternative our data are consistent with a selective transport mechanism. Axonal targeting of GABA B R1a is not altered by its cytosolic C-terminal. Thus, transport may be controlled by a specific GABA B R1a N-terminal adaptor or by a general ER adaptor complex. Identification of these molecules in future studies is needed to strengthen this hypothesis. A mechanism compatible with (i) has been described for the rotavirus VP7 glycoprotein, which is retained in the ER. VP7 is still transported to the axon after BFA treatment [35]. Similar to GABA B R1a, VP7 uses a Golgi-independent intracellular sorting mechanism to reach the axon. One may envision the participation of chaperones in ER transport and targeting. For example, Hsc70 may provide a mechanism to release kinesin from cargo in specific subcellular domains, thereby producing the delivery of axonally transported cargo [36].

Kinesin-1-dependent ER transport of GABA B R1a in axons
Since GABA B R1 is transported along the ER and has a limited residency period within the organelle under physiological conditions we refer to it as an ER-boarded protein. The precise transport mechanism of ER-boarded GABA B R1a in the axon is not clear. A kinesin-1-dependent transport of ER proteins has been described in dendrites [37]. Likewise, a microtubuledependent transport of two ER resident proteins in the axon, GFP-SERCA and GFP-IP 3 R, has been observed in cultured chick dorsal root ganglion neurons [38]. Their bidirectionality and average velocities (, 0.1 mm/s) suggest that the transport is likely non-vesicular, and may represent lateral displacement within the continuous axonal ER membrane or mobility of the organelle itself. Directionality at steady state and average velocities observed for GFP-SERCA and GFP-IP 3 R range between slow (0.001-0.03 mm/s) and fast axonal transport (, 1 mm/s) [38,39]. These intermediate velocities are conserved in our study suggesting that similar mechanisms may operate for the transport of ER-boarded GABA B R1a in axons. A conserved transport mechanism is also supported by the microtubule dependence of axonal GFP-SERCA, GFP-IP 3 R and GABA B R1a mobility. Thus, as reported for ER resident proteins al least one component of the mobility of GABA B R1a is not mediated by simple diffusion. However, precisely how the microtubulecytoskeleton mediates the movement of integral membrane proteins within the continuous ER network is still unclear. According to the mechanisms originally put forward by Tsukita and Ishikawa [40], Waterman-Storer and Salmon [41] and others [42] kinesin-1 may control the lateral displacement of the subunit along the ER membrane in a conveyor-like system or mediate the microtubule-dependent ER sliding of the organelle itself, which may contribute to GABA B R1a transport. Since components of both mechanisms are dependent on kinesin-1 and stable microtubules it is difficult to discriminate between them. Because in this study we wanted to eliminate microtubules that serve as tracks to kinesin-1, the concentrations of nocodazole used were ,15 fold higher than those that affect exclusively the population of dynamic microtubules (100 mM versus , 6 mM). Thus, additionally, we cannot rule out the contribution of dynamic microtubules to the transport of GABA B R1a in the axon [41]. The existence of discrete mobile packets for axonal GABA B R1a also raises the possibility that these are not continuous with the ER network. Electron microscope studies in central and peripheral axons have revealed that the structure of the ER is predominantly contiguous but that it contains occasional free elements [40], [43]. Additionally, isolated and mobile compartments have been observed with fluorescent reporters [38]. Puncta-like behavior may also be explained by transitory tubule fission/fusion events [40] or lateral mobility of protein aggregates. The precise relationship between the mobile structures observed in this and other studies with the continuous ER network requires further studies. Nonetheless, our results support a kinesin-1-mediated axonal transport of ER-boarded GABA B R1a that is in agreement with previous reports.
The majority of the axonal GABA B R1a subunit was static during the live imaging conditions and intervals examined in this study. Additionally, the mobile fraction was higher for a mutant subunit that escapes the ER (MYC-GABA B R1a-EGFP: 19% mobile puncta, n = 259; MYC-GABA B R1a-ASA-AA-EGFP 33% mobile puncta, n = 247). It will be interesting to understand the significance and properties of the static component of GABA B R1a in the ER, and whether ER immobility plays a role in processing or trafficking.
Previous studies have demonstrated that GABA B Rs are segregated intracellularly and that blockade of ER exit results in the accumulation of heteromers in the soma and dendrites of hippocampal neurons [4]. These results suggest that GABA B R subunits are transported into dendrites independently and not as assembled heteromers. They also suggest that newly synthesized GABA B Rs assemble in the ER and exit throughout the somatodendritic compartment prior to insertion at the plasma membrane. This idea is in agreement with transport within the ER and disagrees with a long-haul post-Golgi vesicular transport of GABA B Rs, suggesting a non-canonical trafficking modality for GABA B Rs in dendrites. Thus, long-range ER transport may underlie both dendritic and axonal targeting of GABA B Rs. Since the local complexity of the ER network influences intracellular trafficking it may be utilized by neurons to control the availability of dendritic and axonal membrane proteins that are spatially restricted [44], [45].

Axonal trafficking and ER/ERGIC recycling
In central and peripheral axons the ER is a continuous threedimensional network of irregular tubules and cisternae [40], [43]. As mentioned above, several studies have demonstrated the localization of ER resident proteins and components of the protein folding and export machineries in the axon [19][20]. More importantly, the discovery of local assembly of COPII components in the axon supports the existence of functional ER exit sites required to process newly synthesized proteins that contribute to axonal outgrowth during the early stages of development [46]. However, direct evidence for local protein trafficking within early biosynthetic organelles in the axon is still lacking. In Drosophila, polarized secretion of the EGFR ligand from photoreceptor neurons includes local processing and secretion in the axon, and both mechanisms are controlled by ER localization [47].
More functional evidence is needed to conclusively demonstrate the role of a local trafficking pathway for GABA B R1a in axons. However, a fraction of GABA B R1a colocalizes and moves in synchrony with the ERGIC. This indicates that a local route involving export and retrieval between early biosynthetic organelles may contribute to GABA B R1a trafficking in the axon. A coat protein I complex (COPI) dependent retrieval mechanism from the cis-Golgi to the ER has been postulated for GABA B R1a [48], [49]. Whether this occurs in the axon and whether the remaining secretory steps operate locally to deliver GABA B R1a to the axonal plasma membrane remains unclear. In any case, the presence of the same RXR-type ER retention motif in GABA B R1a and GABA B R1b suggests that despite determining a rate-limiting step for ER exit, dwell time in the ER is not a major determinant of spatial range and axonal localization of GABA-B R1a [50].
Overall our study demonstrates that ER sorting and local transport are relevant for axonal GABA B R trafficking. It is of great interest to determine to what extent the axonal ER is involved in the trafficking of other neurotransmitter receptors and ion channels that travel long distances, especially in long peripheral nerves.