Preliminary studies (using the targeting of viral proteins in neurons infected with vesicular stomatitis virus or influenza virus) on the polarity of neurons determined that MDCK cells and neurons share common pathways for sorting. In support of this
comparison, studies using replication-defective viruses (HSV and Adv) determined that TfR, LDLR, and pIgR (all transported to the basolateral domain of MDCK cells) were all targeted to the somatodendritic compartment in neurons (Jareb and Banker 1998;
Silverman 2001). Furthermore, mutation of the basolateral targeting signals (Yxxθor NPxY) in these proteins, led to a non-polarized distribution in neurons. In addition, after transport to the basolateral/somatodendritic domain, pIgR was transcytosed to the apical/axonal domain, suggesting that transcytotic routes are also conserved between the two cell types. However, clear differences appeared despite this apparent conservation. In MDCK cells, the removal of basolateral signals results in apical localization, whereas a non-polarized distribution occurs in neurons, suggesting that the signals may be conserved but the mechanisms may differ.
To examine axonal targeting, three representative apical proteins were studied, HA, CD8α,and P75/NGFR. In contrast to the similarities between the two cell types for somatodendritic sorting, all three apical proteins exhibited a non-polarized distribu-tion in neurons. Thus, although polarized epithelial cells have provided an invaluable model for studying neuronal polarity, it is now evident that there is a clear need to focus future studies on neurons. As discussed earlier, recombinantly expressed NgCAM is restricted to the axonal membrane. However, if the neurons were perme-abilized, intracellular NgCAM localization revealed a non-polarized distribution, suggesting that NgCAM is selectively polarized only at the level of the plasma membrane. In contrast, the intracellular distribution of TfR mirrors that of surface receptors and is restricted to the dendritic cytoplasm. Thus, NgCAM is either excluded from the plasma membrane of the dendrites and cell soma or rapidly removed after insertion.
The polarized distribution of the recombinantly expressed metabotropic receptors, mGluR2 and mGlur7, are similar to the intracellular polarity observed for TfR and NgCAM. The mGluR2 is restricted to the somatodendritic compartment, whereas mGluR7 is non-polarized (Stowell and Craig 1999). Like the TfR, the mGluR2 may fail to access the axon, due to the lack of an appropriate targeting signal or the presence of an exclusion signal. The expression of a truncation mutant of mGluR2 revealed that the recombinant mGluR2 could now access the axonal plasma membrane in 65 % of neurons, supporting the removal of an axonal exclusion signal. Although the presence of an axonal exclusion signal could explain the distinct localization of mGluR2 and mGluR7, it cannot result in the specific axonal targeting of molecules such as NgCAM, which are not excluded from dendrites. Curiously, the recombinant expression of a similarly truncated mGluR7 construct in neurons resulted in access to the axonal plasma membrane in only 8 % of neurons. This striking observation supports the existence of an axonal targeting signal within the carboxy-terminus of mGluR7.
Furthermore, when this region was fused to telencephalin to produce a chimaeric hybrid protein, the chimaera was observed in the axonal plasma membranes of 77 % of neurons (0.7 % for wild-type telencephalin). Thus the axonal targeting signal residing in mGluR7 appears to be both necessary and sufficient for axonal targeting. When
this region was added to the mGluR2, the axonal targeting signal dominated over the exclusion signal. However, the targeting of truncated mGluRs were not identical, as expected for the absence of both targeting and exclusion signals, suggesting that other targeting/exclusion signals may function as cryptic (that may not normally function) targeting signals.
Just as similarities and differences exist in the polarized transport of proteins between epithelial cells and neurons (Jareb and Banker 1998), the same appears to be true between different neuronal cell types as well as within an individual neuron, depending on previous activity (Agno et al. 2000). For example, the recombinant expression of mGluR5 in striatal or cerebellar granule neurons (CGNs) revealed that mGluR5 was restricted to the cell soma of CGNs, but was dendritically localized within striatal neurons. As these neurons expressed different homer proteins (mGluR-interacting proteins known to regulate ER-retention and cell surface delivery), it was hypothezised that the lack of homer 1 expression in CGNs might explain the failure to export mGluR5 from the cell soma. Consistent with this hypothesis, when homer 1b was transfected into CGNs it could be detected in dendrites. Furthermore, when co-transfected with mGluR5, both proteins were detected in dendrites, but not axons.
Interestingly, when homer 1a (absent from both striatal and CGNs) was transfected into CGNs, it was found to gain access to both axons and dendrites. Upon co-transfection with mGluR5, both molecules could be detected in axons and dendrites. Similar results were observed for homer 1c in cortical neurons, which caused an increase in the dendritic trafficking of GluR1a. Homer 1a is an immediate early gene product, whose expression is induced after intense depolarization of neurons (including CGNs) and may deliver mGluRs to dendrites and axons during high-level neuronal activity, as occurs during convulsive seizures and the induction of long-term potentiation (LTP)(see Chapters 14 and 15). Homer proteins do not appear to be essential, as endogenous mGluR1a is dendritically expressed in CGNs, despite the lack of these proteins.
Two of the greatest advances in the study of neuronal trafficking must be the discov-ery of GFP and the ability to transfect neurons. When used together, it is possible to follow neuronal trafficking pathways, in real time, in living neurons. Apart from the breathtaking video images generated of protein trafficking within neurons (for exam-ple see: http:/www.neuron.org/cgi/content /full/26/2/465/DCI), significant scientific advances are being made. In the study of Burack et al. (2000), TfR-GFP and NgCAM were observed to traffic within discrete organelles. These organelles are highly pleo-morphic and dynamic, ranging from small vesicles to long (>1 µm) tubules. Movement for both proteins occurred bi-directionally in dendrites (TfR and NgCAM) and axons (NgCAM).
The trafficking of transport vesicles in both dendrites and axons was inhibited by >80 % following treatment with the microtubule-disrupting agent, nocodazole, suggesting that transport occurs along microtubule tracks in both types of neurite.
Microtubule tracks in dendrites exist in both orientations, with their plus and minus
ends facing away from the cell body. In contrast, microtubules in axons exhibit a uniform polarity, with their plus ends facing away from the cell body (Fig. 4.2). If car-rier vesicles containing dendritic proteins, such as TfR, exclusively associate with minus end-directed microtubule motors (e.g. dynein) then they would be excluded from axons, yet be capable of bi-directional transport in dendrites. Conversely, axonal vesicles, containing molecules such as NgCAM, could associate with plus end-directed motors (e.g. kinesins), gaining them bi-directional access to dendrites, but only uni-directional (anterograde) access to axons. Although this could explain the polarized distribution of proteins discussed above, it is not consistent with the real time studies identifying bi-directional trafficking of both dendritic and axonal transport organelles.
When the chimaeric molecules, amyloid precursor protein (APP)-YFP and synapto-physin (p38)-GFP, were recombinantly expressed in hippocampal neurons in culture and their transport along axons investigated by two-colour video microscopy, it was found that these proteins were segregated prior to transport along axons (Kaether 2000). In doubly transfected neurons, APP-YFP and p38-GFP exhibited mutually exclusive distributions. APP-YFP was transported rapidly (~4.5 µmS1) within elon-gated tubules up to 10 µm in length. In contrast, p38-GFP was restricted to slow (~1 µmS1) tubulovesicular carriers. The use of antisense oligonucleotides to block the translation of Kinesin mRNA, resulted in a disruption of APP-YFP carriers but did
Fig. 4.2 Transport into dendrites and axons occurs along microtubules. Microtubules in dendrites are oriented in both directions, whereas the microtubules in axons are all polarized with their plus end distal to the cell body. Vesicles carrying cargo into dendrites are excluded from axons, but axonally targeted vesicles are transported into both axons and dendrites.
Thus, dendritic versus axonal sorting may occur early in the biosynthetic pathway, with axonal targeting being achieved by a retention mechanism or the failure of vesicle to fuse with dendritic plasma membrane. Dendritic targeting may be achieved by exclusion from the axon.
Dendrite
Axon
mGluR7/NgCAM mGluR2/TfR
not affect p38-GFP transport. Thus, other microtubule motors may be involved in some of these transport pathways, particularly within the axon (Diaz-Nido and Avila, this volume, Chapter 15).