The data in this study indicate that Myosin VI plays a role in the enrichment of transmembrane proteins at the axonal surface. Myosin VI facilitates the relative enrichment of proteins at the surface of the axon by increasing the rate of endocytosis within the somatodendritic compartment relative to the axonal compartment. Overall, our results are consistent with a model whereby axonal proteins are transported by kinesin motors to both the axon and the dendrites following release from the Golgi apparatus (Figure S14
). Following arrival of protein at the somatodendritic plasma membrane, it is endocytosed through the actions of Myosin VI. This participation of Myosin VI in dendritic endocytosis of axonal proteins is consistent with numerous reports documenting its role in endocytosis in epithelial cells and neurons as well as in nonpolarized cells 
. It remains to be determined the exact mechanism whereby specificity of Myosin VI action is achieved. It is possible that Myosin VI interacts either directly or indirectly with axonal proteins, or, alternatively, it is possible that Myosin VI interacts with the endocytic machinery associated with vesicles carrying these proteins 
. In the future it will be important to fully explore this question.
With the results of this study, there is now evidence that actin and myosin play pivotal roles in the concentration of proteins on the surface of either the axon or the dendrites; however, the role of microtubule-based mechanisms in these processes is less clear. A number of observations indicate that kinesin motors might be involved in steering vesicles to polarized compartments. For instance, the motor domain of Kif17, which transports dendritic proteins 
, carries vesicles into the axon less efficiently than Kif5, which transports both axonal and dendritic proteins 
. Tailless Kif5 targets to axons in a manner that is dependent on microtubule dynamics 
, and blocking Kif5 function with dominant negative constructs disrupts neuronal polarity 
. Additionally, tyrosination, a phenomenon that occurs in dendrites, but not axons, causes Kif5 to bind at lower affinity to dendritic microtubules in comparison to those present in the axon 
. Finally, Kif5 prefers to bind to stable microtubules, which predominate in the axon, whereas other kinesins can also interact with unstable microtubules, which predominate in dendrites 
. Although these experiments suggest that Kif5 has a bias towards axonal transport, it is also capable of carrying proteins such as GluR2 to the dendrites 
. The work in this paper and our previous work would suggest that APP, which is also carried by Kif5 
, and GluR2 localize differently because APP is carried in vesicles that associate with active Myosin VI, whereas vesicles carrying GluR2 likely are influenced by the actions of a plus-end-directed Myosin, such as Myosin Va 
. Nonetheless, our experiments do not rule out an active role for kinesins in polarized targeting. For instance, it has been shown that Kif5 likely plays a role in the initial establishment of neuronal polarity during early development in vitro when it carries C-Jun N-Terminal Kinase to the nascent axon 
. Thus, under certain circumstances, kinesin motors are capable of targeting specifically to one or the other polarized compartment. However, the results in this paper and our previous study 
would indicate that in many cases contributions by myosin motors are required for polarized trafficking.
In addition to kinesins, the dynein complex has also been suggested to participate in the polarized targeting of dendritic proteins, since causing organelles to link to dynein results in their transport specifically to dendrites 
. While this effect is remarkably robust, it is not surprising given the fact that only dendrites contain microtubules that are oriented in a direction that permits movement of dynein away from the cell body and into the process 
. Although blocking dynein with Dynamitin blocked dendritic targeting of GluR2, this disruption was not distinguished from the disruption of neuronal polarity in general. In contrast, blocking the function of Myosin Va or Myosin VI specifically disrupts the polarized distribution of subsets of proteins without destroying the overall polarity of the cell (
; and S6
–S8). Furthermore, the assertion that dendritic proteins are specifically transported by dynein motors and not by kinesins would appear to contradict numerous studies 
. A model where dynein plays a crucial role in the maintenance of the cytoskeletal structure that is essential for neuronal polarity, while myosin motors in combination with kinesins mediate trafficking, is consistent with both our work and that of others.
Recent experiments support the existence of a vesicle filter dependent on actin within the axon initial segment. For instance, when large molecules are injected into the cell body of dissociated hippocampal neurons in culture, they are excluded from the axon in control neurons, but not in neurons exposed to Cytochalasin D, which promotes actin depolymerization 
. Latrunculin B, which also promotes actin depolymerization, disrupts the localization of numerous polarized proteins including NgCAM 
. Moreover, electron microscopy studies show the presence of actin filaments, albeit of indeterminate orientation, below the axolemma 
. Our previous results suggested that this filter might play a role either in preventing entry of dendritic proteins into the axon 
or in promoting their entry into dendrites. In this study we found that a protein bound to Myosin VI (CD8-MVIBD) is localized intracellularly to both the axon and the dendrites (Figure S9
), which would imply that there is nothing preventing it from entering the dendrites and thus that no vesicle filter exists in the proximal dendrites. In addition, a vesicle filter present in the axon initial segment that prevents proteins associated with Myosin Va from entering the axon would have the opposite effect on proteins associated with Myosin VI. Thus, Myosin VI might interact with actin filaments to guide vesicles containing axonal proteins and carried by kinesins towards the axon (Figure S14
The presence of a vesicle filter in only one compartment could explain why axonal and dendritic proteins are trafficked by distinct mechanisms in mammalian neurons: (1) Axonal proteins associated with Myosin VI initially enter both compartments, but are concentrated in the axon in part through Myosin VI-dependent endocytosis from the dendritic plasma membrane. (2) Vesicles carrying dendritic proteins, which associate with Myosin Va, are prevented from entering the distal axon by its vesicle filter, which causes the vesicles to be trafficked directly to the dendrites. In invertebrates, however, it appears that localization of both axonal and dendritic proteins depends on endocytosis from the plasma membrane of the opposite compartment. Thus, in Caenorhabditis elegans
, several dendritic receptors including the acetylcholine receptor and a glutamate receptor are localized to dendritic membranes through axon-specific endocytosis, while Synaptogyrin is localized to the axon through endocytosis from the somatodendritic membrane 
. It seems likely that in the relatively compact neurons of invertebrates, endocytic mechanisms alone are sufficient to localize transmembrane proteins efficiently to the membrane of either the axon or the dendrite. Thus, it is tempting to speculate that the advent in vertebrates of neurons with very long, thin axons dramatically increased the cost of localization strategies dependent on endocytosis. This development might have exerted sufficient selective pressure to facilitate the formation of a cytoskeletal structure, the vesicle filter, which prevents vesicles containing dendritic proteins from entering axons and promotes the entry of vesicles containing axonal proteins.