4.7.1 mRNA targeting
Until recently, it has been envisaged that protein localization is achieved exclusively by protein trafficking events preceding anchoring mechanisms. However, a completely different strategy, involving the targeted transport of mRNA molecules is gaining popularity (reviewed by Job and Eberwine 2001). A particular attraction of this mecha-nism is its economical use of cellular components, since one mRNA molecule has the capacity to produce many molecules of the protein product. The indication of localized dendritic translation of mRNA was obtained by electron microscopy by the observation of polysomes at the base of dendritic spines. Individual mRNA species (encoding MAP2, CaMKII, and glycine receptors) were subsequently identified by in situ hybridization.
The full potential for localized translation was realized by the extraction of mRNA from mechanically isolated dendrites using a patch pipette, followed by PCR amplifica-tion. Using this methodology, it was determined that, of the 10 000 mRNAs thought to be present in neuronal soma, approximately 400 were estimated to be present in dendrites. More significantly, differential display revealed that all dendrites do not contain the same complement of mRNAs, raising the possibility of differential target-ing, or at least, specificity.
To date, mRNAs for structural proteins (e.g. MAP2), enzymes (e.g. CamKII), growth factors (e.g. BDNF), ligand-gated ion channels (e.g. glycine receptors), and transcrip-tion factors (e.g. CREB) have been found within dendrites. The significance of mRNA being present within dendrites was enhanced by the immunocytochemical colocaliza-tion of proteins involved in the translacolocaliza-tion process and present within the ER and Golgi, a prerequisite for the assembly and trafficking of integral membrane proteins such as ion channels (Gardiol et al. 1999).
Direct evidence that dendritically localized mRNA are actually translated locally was first obtained via the incorporation of radiolabelled amino acids into isolated dendrites. Synaptoneurosome preparations were used to detect and quantify protein synthesis. Intriguingly, a glutamate-responsive mRNA encoding an RNA-binding protein, FMR1 (Fragile-X mental retardation protein), which can regulate translation from polysomes was detected. Thus, synaptic activity may initiate local protein synthe-sis, ensuring a supply of required proteins exclusively to the site(s) requiring them.
Definitive proof that dendrites do possess the capacity to translate proteins came from an elegant molecular approach. Dendrites were isolated from neuronal cell bod-ies and transfected with a recombinantly transcribed mRNA. Consistent with dendritic translation being regulated, expression could only be detected following stimulation with growth factors (e.g. BDNF).
Given the almost infinite number of distinct synapses, it is unlikely that distinct tar-geting signals could exist for each. Alternatively, dendritically targeted mRNAs could be targeted to all synapses, but translated only at synaptically active sites. Combined with temporally regulated transcription in the nucleus, a simple, yet effective, mecha-nism for selective targeting could be achieved. As synaptic activity is known to regulate the transcription of certain genes in the nucleus, it is tempting to speculate that a mechanism could exist in which active synapses might specify the proteins that need to be recruited. Support for such an hypothesis has been provided by the discovery of the transcription factor (CREB) mRNA and protein within dendrites. More importantly, the CREB protein could be activated locally by phosphorylation (a requirement for transcription) and was capable of retrograde transport to the cell nucleus. Thus, restricted synaptic activity may initiate the local translation of proteins required within the synapse as well as those required in the nucleus for the transcription of new mRNAs. An exciting possibility is whether certain types of synaptic activity could lead to the transcription of specific genes, i.e. the provision of a shopping list, rather than just requesting ‘food’, for delivery. Of course, what is delivered by such a mechanism is the recipe, in the form of mRNA molecules.
4.7.2 Protein targeting via lipid rafts
It is becoming increasingly apparent that the segregation of specific proteins into sphingolipid/cholesterol-containing microdomains is a commonly used sorting mech-anism operating within the exocytic and endocytic pathways (reviewed in Ikonen 2001). The role of lipid rafts in organizing the exocytotic machinery has been impli-cated by their association with SNARE proteins. The association of VAMP2 (a vesicular (v)-SNARE) with lubrol-insoluble lipid domains, such as are found on synaptic vesi-cles, suggests a role of lipid rafts in targeting. Similarly, the association of syntaxin 1A (a target (t)-SNARE) within a distinct subset of lipid rafts (based on detergent solubil-ity characteristics) suggests that sites of exocytosis may be restricted to discrete sub-domains in the plasma membrane. Although VAMP2 and syntaxin 1A are involved in the regulated exocytosis of presynaptic vesicles, syntaxin 1A and SNAP-25 also occur along the axonal plasma membrane, implicating a more general role for lipid rafts in exocytosis. Interestingly, different Kv channel isoforms show distinct localization pro-files with respect to lipid rafts, with Kv1.5 in caveolar rafts and Kv2.1 in non-caveolar rafts, while Kv4.2 is not associated with any lipid rafts. Thus, distinct lipid rafts may play a critical role in determining the sites of exocytosis for different cargo proteins.
Furthermore, these specialist sites may not be restricted to axonal destinations. AMPA receptors have been found to be equally distributed between lipid rafts and postsynap-tic densities and may be recruited by an interaction with raft-associated GRIP, via the AMPA receptor GluR2 subunit. In addition, the α7 nicotinic acetylcholine receptor localization and clustering in somatic spines of ciliary neurons requires lipid raft asso-ciation. The full significance and potential of lipid rafts in protein sorting remains to be established (reviewed in Tsui-Pierchala et al. 2002).
4.7.3 Specific transport proteins and pathways
With the advent of 2-hybrid (yeast, bacterial and mammalian) screening, a molecular approach to the identification and cloning of interacting proteins, has come the identi-fication of numerous clustering, anchoring, signaling, and transporting proteins involved in postsynaptic receptor function. This technology has been particularly fruitful in the identification of participants in the transport of GABAAand AMPA receptors to the cell surface (Kittler and Moss 2001; Sheng and Lee 2001; Malinow and Malenka 2002). Equally important has been the development of methods to transfect neurons and the application of GFP chimaeras to the study of trafficking pathways leading to synaptic targeting.
4.7.3.1 GABAA receptors
It is known, from transgenic studies, that GABAAreceptor synaptic targeting at most inhibitory synapses requires both the γsubunit and gephyrin. GABARAP (GABAA receptor associated protein) was the first yeast two-hybrid hit for the GABAAreceptors.
GABARAP interacts with the γ2subunit of GABAAreceptors, but is rarely associated with clustered GABAAreceptors or gephyrin. Instead, GABARAP appears to be con-centrated at transport sites on the Golgi, where it interacts with N-ethylmaleimide-sensitive fusion protein (NSF), suggesting a role in GABAA receptor trafficking. This is supported by the close structural homology of GABARAP to GATE-16 (Golgi associ-ated ATP enhancer) a known transport molecule. The crystal structure of GABARAP suggests the ability to bind the γ2subunit and tubulin on opposite faces. Thus, GABARAP may select vesicles (possibly even sort the receptors into vesicles) with γ2 -containing receptors and transport them along microtubules to inhibitory synapses.
As GABARAP is not localized to these sites, it presumably hands over the receptors to microtubule-associated gephyrin clusters and returns to the Golgi (reviewed in Kneussel 2002).
Using the large intracellular loop of the α1subunit of GABAAreceptor as bait in a 2-hybrid screen revealed Plic-1 as an interacting protein capable of binding to all αand βsubunits, but not γor δ. Like GABARAP, Plic-1 is found on intracellular membranes and is not found significantly at synapses. Having identified the GABAA receptor-binding domain, the authors delivered (using the antennapedia internaliza-tion peptide) inhibitory peptide to cells. These experiments revealed a requirement for Plic-1 in the maintenance of GABAAreceptors at the cell surface. As GABAAreceptors constitutively recycle in neurons, Plic-1 may be required for recycling to the plasma membrane, or protection from degradation during recycling. Plic-1 is a ubiquitin-like protein and may be a negative regulator of the proteasome. It is not clear if the sole function of Plic-1 is to prevent GABAAreceptor degradation, thus leaving more recep-tors capable of recycling back to the cell surface, or if Plic-1 may play a direct role in returning receptors. Plic-1 may also increase receptor number by protecting receptors from ubiquitin-dependent degradation within the ER during assembly. Unlike
GABARAP, Plic-1 would be capable of maintaining both synaptic and extrasynaptic surface GABAAreceptors.
As mentioned above, multiple sorting stations may exist in neurons, serving to filter and direct proteins to their final destination. In addition to providing specificity, these compartments may also offer the opportunity to regulate the delivery of its cargo to the cell surface. Indeed, the exocytosis of synaptically targeted (γsubunit-containing) GABAAreceptors appears to be regulated by a mechanism distinct from extrasynaptic (lacking γsubunits) GABAAreceptors. In recombinant expression systems, both αβand αβγreceptors constitutively recycle between the cell surface and peripheral endosomes. However,αβγreceptors are subsequently diverted from peripheral early endosomes and routed into a perinuclear (in fibroblasts) late endosomal compartment.
Upon PKC stimulation, exocytosis from this late compartment is blocked, leading to a reduction in the surface expression ofαβγ, but not αβreceptors. Such a mechanism may provide the basis for the regulation of GABAAreceptor synaptic targeting (involv-ing GABARAP?) (Kittler and Moss 2001; Kneussel 2002). The presence of a subcellular compartment from which GABAAreceptors may be recruited rapidly is supported by the findings that insulin (via tyrosine kinase activity) causes the fast (<10 min) translocation of GABAAreceptors to the cell surface of transfected fibroblasts or synapses in neurons. The speed of surface delivery is consistent with receptor recruit-ment from an endosome. Thus, kinase activity is capable of modulating the cell surface delivery of GABAAreceptors by either inhibiting (PKC) or promoting (tyrosine kinase) receptor targeting from endosomes.
4.7.3.2 NMDA receptors
The delivery of NMDA receptors to the plasma membrane is also subject to kinase regulation. In contrast to the findings for GABAAreceptors, PKC activation was found to promote NMDA receptor targeting to the cell surface. Importantly, endogenous levels of PKC stimulation (via mGluR1 activation) are sufficient to increase surface tar-geting of NMDA receptors. This tartar-geting can be inhibited by botulinum neurotoxin A and a dominant negative mutation of soluble NSF-associated protein (SNAP-25), implicating the involvement of SNAP-25 dependent exocytosis. This PKC-dependent pathway is conserved in Xenopus oocytes, fibroblasts, and hippocampal neurons.
Intriguingly, NMDA receptors are also recruited to the cell surface of Xenopus oocytes in response to insulin treatment, via a SNAP-25 dependent process. It appears that the regulated exocytosis of NMDA receptors may also occur in response to synaptic activ-ity. In the adult (but not neonate; see below) CA1 region of the rat hippocampus, an increase in synaptic strength leads to the rapid surface expression of NMDA receptors by a PKC and tyrosine kinase-dependent mechanism (Grosshans et al. 2001). Given the opposite nature, and distinct localization of GABAAand NMDA receptors in the adult, it seems likely that at least two distinct, highly homologous, sub-synaptic endosomal sorting compartments co-exist within neuronal dendrites.
4.7.3.3 AMPA receptors
In neonates, NMDA receptors are relatively fixed components of the postsynaptic den-sities, whereas AMPA receptors are more loosely associated. Long-term potentiation (LTP) and long-term depression (LTD) are modifications to synaptic strength, expressed as AMPA receptor responsiveness to glutamate, and are currently the best molecular correlate of learning and memory. The modulation of synaptic strength has been perceived traditionally to reflect alterations in neurotransmitter release. However, the modulation of AMPA receptor numbers in postsynaptic membranes might also provide a powerful mechanism to modulate synaptic strength. Indeed, a large body of evidence supporting this hypothesis has emerged over the last three years (reviewed in Sheng and Lee 2001; Malinow and Malenka 2002).
The identification of a postsynaptic compartment from which regulated exocytosis might occur was achieved by loading cells with the membrane dye FM1-43. Prolonged labelling was necessary for the dye to access a compartment from which regulated exo-cytosis could occur, consistent with its distribution within the TGN/late endosome.
The exocytotic release of FM1-43 could be induced by glutamate, electrical stimulation (50Hz), calcium and CaMKII stimulation, conditions necessary for the expression of LTP. In primary hippocampal neurons, this calcium-evoked dendritic exocytosis is only evident after 7 days in culture, unless αCaMKII (absent until day 7) was expressed using recombinant viral constructs. The requirement for postsynaptic membrane fusion was confirmed by the postsynaptic introduction (via a microelectrode) of several agents (N-ethylmaleimide, NSF-SNAP binding inhibitory peptides and botu-linum toxin) that block membrane fusion events. In all cases, LTP in the CA1 region of the hippocampus was blocked, confirming the importance of postsynaptic membrane fusion in the expression of LTP at these synapses.
NSF plays a critical role in vesicle fusion with target membranes in both constitutive and regulated exocytosis. NSF is recruited to v-SNARE/t-SNARE/SNAP-25 complexes via soluble NSF-attachment proteins. It was quite unexpected, therefore, when 2-hybrid screening identified a direct interaction between the AMPA receptor GluR2 subunit and NSF. Support for the validity of this interaction was obtained by the perturbation of NSF-GluR2 interactions. The injection of inhibitory peptides (corre-sponding to NSF-binding site on GluR2) or monoclonal antibodies (against NSF) into postsynaptic hippocampal neurons resulted in a rapid rundown of AMPA receptor currents. Although GluR2 is clearly not a SNARE protein, it is possible that NSF may mediate conformational changes in GluR2 exposing a plasma membrane targeting signal or dissociation from an anchoring protein, such as GRIP, to initiate endocytosis.
An exciting piece of evidence in AMPA receptor trafficking was obtained from stargazer mice. The main cerebellar defect is found in cerebellar granule neurons and is expressed as a complete lack of synaptic AMPA receptors. The protein encoded by the stargazer gene, stargazin, is able to bind both AMPA receptors and PSD-95. Importantly, the lack of surface and synaptic AMPA receptors could be restored by the recombinant
expression of wild-type stargazin in cultured neurons from these mice. In contrast, when these neurons were transfected with a mutant stargazin lacking the PDZ-binding domain, both the PSD-95 interaction and synaptic clustering were not observed.
However, this mutant could still interact with, and deliver, AMPA receptors to the cell surface. This observation led to the discovery of the dual role performed by stargazin;
to deliver AMPA receptors to the cell surface (at extrasynaptic sites) and following their subsequent lateral diffusion into the synapse, to hand over its charge to PSD-95 (Schnell et al. 2002).
The possibility of AMPA receptor recruitment as being the site of expression of LTP was spurred on by the identification of silent synapses. At resting membrane poten-tials, when NMDA receptors are blocked by Mg2+, only AMPA receptors respond to glutamate. In the absence of functional AMPA receptors, glutamatergic synapses are therefore silent. The majority of synapses are silent at early stages of development but silent synapses are uncommon in the adult brain. A morphological correlate of silent synapses, using the immunohistochemical localization of GluR1, suggested that silent synapses may lack AMPA receptors.
Conclusive evidence that postsynaptic AMPA receptor recruitment does occur during LTP was provided by an investigation into the trafficking of recombinant GluR1-GFP. GluR1-GFP was introduced into neurons in organotypic hippocampal slice cultures by a Sindbis virus expression system. The distribution of GluR1-GFP was determined directly by GFP fluorescence and surface receptors identified using GFP antibodies in non-permeabilized cells. As for endogenous GluR1, the GluR1-GFP is predominantly intracellular. However, upon brief tetanic stimulation (LTP-inducing), GluR1-GFP is recruited to both silent and active synapses.
In keeping with a role for CaMKII in regulated exocytosis and LTP, when a constitu-tively active form of CaMKII (tCaMKII-GFP) is expressed in neurons, enhanced synaptic transmission and GluR1-GFP recruitment occurs. GluR1-GFP recruitment required its PDZ-binding domain implying that GluR1 recruitment as well as cluster-ing (Cattabeni, this volume Chapter 7) requires PDZ domain proteins.
The accumulation of recombinantly expressed GluR2 requires an association with NSF, ABP (AMPA receptor-binding protein), and/or GRIP (glutamate receptor inter-acting protein). In contrast to the findings for GluR1, GluR2-GFP was constitutively expressed at synapses by an activity-independent mechanism. A careful examination of the synaptic targeting of recombinantly expressed GluR1 versus GluR2 revealed that GluR1 is retained intracellularly until recruited to active or silent synapses upon acti-vation, whereas GluR2 receptors appear to be excluded from silent synapses. However, endogenous AMPA receptors are predominantly composed of both GluR1 and GluR2 subunits. The recombinant expression of GluR1/2 receptors revealed a trafficking itin-erary indistinguishable from recombinant GluR1 homomeric receptors. New GluR1 or GluR1/2 receptors appear transiently as micropuncta in non-synaptic areas prior to their accumulation at synapses. In contrast, GluR2 appear to be delivered directly, and rapidly, to synapses. The initially non-synaptic targeting of GluR1-containing
receptors may be explained in one of two ways. Firstly, GluR1-containing receptors might diffuse laterally within the plasma membrane (Borgdorff and Choquet 2002) until ‘captured’ by synaptic proteins (Schnell et al. 2002). Alternatively, it is possible that synaptic localization may be achieved by random surface delivery, followed by receptor concentration into clathrin-coated vesicles, transport into a sorting endosome and subsequent targeting to synapses.
By performing these recombinant studies it has been possible to tease apart the relative subunit contributions to endogenous GluR1/2 receptor targeting. The most likely value of these differential targeting mechanisms is that GluR1 might be required for activity-dependent plasticity that recruits new receptors to silent or active synapses.
Synaptic potentiation achieved in this way, may be maintained by the constitutive re-delivery of recycling receptors, by virtue of GluR2-dependent interactions. With the development of these molecular biological strategies to the study of LTP, has come the shift in popularity from the expression of LTP being exclusively presynaptic to predominantly postsynaptic. However, the pendulum of opinion may soon return to neutrality, with the discovery of an increased targeting of the presynaptic vesicle-associated protein, synaptophysin, to synapses during LTP (Antonova et al. 2001).
Thus, the enhancement of synaptic strength during LTP may be achieved by either an increase in neurotransmitter release, or a concomitant increase in AMPA receptor numbers, or both. Of course, increasing either of these components would be futile if they are already present in excess. Variability between synapses of different brain regions may account for the discrepancies observed in the relative contributions of presynaptic and postsynaptic components to LTP.