In neurons, the microtubule cytoskeleton is organized differently in dendrites and axons. Dendritic microtubules have mixed polarity i.e. plus-ends can face either away from or towards the cell body, whereas the microtubules in the axon have a uniform polarity with all plus-ends facing outwards 3637 (Figure 1). Recently, by imaging microtubules using fluorescently tagged α-tubulin and microtubule plus-end binding proteins it was demonstrated that neuronal microtubules are very dynamic and regularly enter dendritic spines 3839. Microtubules grow with their plus-ends directed towards the spine and are able to repeatedly enter the same spine 39 (Figure 1). In contrast, microtubule nucleation within spines and subsequent polymerization into the dendrite shaft has not been observed. Therefore, polarized growth of microtubules into spines could facilitate delivery of synaptic cargo directly to the postsynaptic site (Figure 1). Although microtubule-dependent cargo transport in dendritic spines has not been directly visualized yet, it will be interesting to investigate whether the translocation of excitatory postsynaptic proteins or other cellular components, such as organelles and RNA particles (RNPs), rely on microtubule based transport 40414243 (Table 1). On the other hand, several studies directly demonstrate the importance of the actin-based transport mechanisms at excitatory synapses. Myosin V, for example, mediates RNP translocation into dendritic spines 34, is required for the spine localization of smooth endoplasmic reticulum at Purkinje cell synapses 44 and transports AMPA receptor containing recycling endosomes during synaptic potentiation 4546.

One way to directly influence synaptic cargo transport is to generate functional diversity by modifying the cytoskeleton. Motor proteins recognize these spatial cues and establish specific synaptic cargo trafficking routes. It has recently been demonstrated that post-translational modification (PTM) of microtubules can alter their stability and motor protein binding characteristics 47. Microtubules can acquire a variety of evolutionarily conserved PTMs including polyglutamylation, polyglycylation, detyrosination, acetylation, phosphorylation and palmitoylation 48. In most cases, modifying enzymes act preferentially on tubulin subunits already incorporated into microtubules. It is tempting to speculate that, due to the specific organization of microtubules in neurons, microtubule PTM is of even greater importance than in non-neuronal cells. Recently, it was found that polyglutamylation targets kinesin-3 family member KIF1A motors 49 and tyrosination recruits kinesin-1 family member KIF5 motor proteins 50. Now, it will be critical to identify the forward and reverse enzymes responsible for the various modifications of neuronal microtubules.

The motor domains responsible for motility and force generation are connected by a stalk to the tail domain, which directly or indirectly binds to their cargo 12. Another possible mode of transport regulation is direct modulation of motor protein activity. Tremendous progress has been made in the mechanistic understanding of molecular motors - both in force generation and directional movement 12325152. For instance, a recent study shows that the specific structure of dynein, with six AAA+ ATPase domains, a linker and a microtubule binding stalk, explain the characteristic dynein stepping pattern 53. The stepping behavior of kinesin was shown to be regulated by intramolecular strain, allowing the kinesin motor to maintain its plus-end directed movement and also coordinate the two motor domains 54. Moreover, several studies demonstrate that myosin and kinesin motors can fold back on themselves in order to regulate their own motor activity 39. In the absence of bound cargo, the tail domain interacts with the motor domain and inhibits its activity 5556. Furthermore, motor accessory proteins such as dynactin have been shown to increase the processivity of dynein and kinesin-2 575859.

Another intriguing aspect of transport regulation lies in the use of multiple motors by a single cargo (Figure 1). Various synaptic cargos were shown to move bidirectionally along the microtubule network and some can even switch between actin and microtubules (Table 1). Several models have been proposed to explain bidirectional transport. The "tug-of-war" and "coordination" models both describe a situation where different types of motors are simultaneously bound to one cargo 20. In the "tug-of-war" model, motors of both minus-end and plus-end directionality will create a force that is decisive for the direction of cargo transport. The back and forth movement of synaptic cargo can be explained by motor loss or changes in activities of motors, causing the opposing motors to take the lead 60. The "coordination" model predicts that when one motor is actively engaged on the microtubule, the opposing motor is turned off. Here, disruption of one set of motors causes transport defects in both directions 61. Both transport machineries are most likely controlled by large regulatory complexes present on the cargo, consisting of Rab GTPases, scaffolding and signaling proteins 1020. In this way, signaling proteins can turn one of the motors "on" and the opposing motor "off" and thus coordinate the activity of both motors. It was known from in vitro studies that cargos that contain multiple motors were able to move further and faster than cargos which engaged only one motor 6263. However, a recent study using Drosophila embryos showed that one cargo that engages multiple motors in fact does not travel faster nor does it travel a greater distance 64. These data suggest that in vivo regulation of motor proteins requires additional factors, possibly in the form of motor-adaptor complexes on moving cargo.

The most direct mechanism of membrane association is attachment to the phospholipid bilayer. Myosin I motors that possess a basic tail domain bind to acidic phospholipids 65 and kinesin-3 family members bind to lipids via their pleckstrin homology domain in the tail region 66. Dynein cannot directly bind to protein-free liposomes, it requires dynactin and spectrin to link to acidic phospholipids in the membrane 5967.

Motor proteins can interact with receptors or integral membrane proteins on membrane cargo vesicles. Kinectin is a membrane protein in the endoplasmic reticulum and was the first proposed KIF5 receptor to anchor motors on cargo membranes 6869. KIF5 has been suggested to bind directly to amyloid precursor protein (APP), a transmembrane protein of which proteolytic fragments give rise to amyloid plaques in Alzheimer's disease patients 70, although some other studies were unable to find evidence for a direct interaction 71. In photoreceptor cells, the light chain of dynein, Tctex-1, directly binds to the G-protein coupled receptor rhodopsin 72. Rhodopsin mutations responsible for retinitis pigmentosa inhibit the interaction with dynein. Tctex-1 also interacts with neurotrophin Trk receptors, suggesting that the dynein motor complex is responsible for retrograde transport of Trk receptors 73. Consistently, real-time imaging revealed that dynein is required for rapid transport of internalized, activated Trk receptors from axon terminals to the cell bodies 74.

Synaptic cargos engage multiple types of motors for both actin and microtubule based transport. There is strong evidence that the cooperation of multiple motors on a single cargo is important to regulate cargo transport 19. Communication between different types of motor on a cargo is most easily envisaged if motors are in direct physical contact. Myosin V and KIF5 have been shown to bind directly and enhance each other's processivity 7576. This motor-motor interaction may represent a mechanism by which the transition of vesicles from microtubules to actin filaments is regulated. Moreover, physical interactions between dynein and kinesin motor proteins have also been reported - direct binding between cytoplasmic dynein and KIF5 77 and Xenopus kinesin-2 associated protein (XKAP) and dynactin has been described 78. Motor-motor associations may also be a mechanism to coordinate anterograde and retrograde cargo movements along microtubules.

Motor proteins can also attach to cargo through binding of signaling proteins and associated components. Jun N-terminal kinase (JNK) interacting proteins (JIPs) are of key importance for axonal cargo trafficking 79 - in Drosophila, JIP mutants cause aberrant accumulation of axonal cargos, closely resembling the phenotype of kinesin-1/KIF5 mutants. JIP functions as an adaptor protein by binding directly to kinesin light chain and carries Reelin receptor ApoER2-containing vesicles 80. Moreover, activation of the JNK signaling pathway triggers cargo release 81.

Both KIF5 and dynein are the primary motors for mitochondria trafficking in Drosophila motor axons 82. The Miro GTPases have recently been demonstrated to anchor in the outer mitochondrial membrane and regulate mitochondrial transport and synaptic function 9. Mitochondria coated with Miro recruit KIF5 through the association with Milton/TRAK proteins 83. In flies, loss of Miro and Milton function display similar phenotypes - abnormal clustering of mitochondria in the cell body and impaired synaptic function at the neuromuscular junction 8485.

Synaptic scaffolds are multifunctional proteins with several protein-protein interaction modules 137. Several studies demonstrate that scaffolding proteins bind directly to motors and function as motor-cargo linkers 108687. The motor-cargo complex on glutamate receptor-containing vesicles consists of glutamate receptor subunits, scaffolding proteins and kinesin motors 8889. The NMDA receptor NR2B subunit interacts with KIF17 through a scaffolding protein complex that consists of Lin7 (also known as MALS/Velis), Lin2 (CASK) and Lin10 (Mint1). Consistently, NR2B and KIF17 have been localized on the same vesicle and move within the dendritic shaft 90. A similar motor-cargo model can be made for AMPA receptor transport - AMPA receptor subunits GluA2/3 (also named GluR2/3) interact with KIF5 through the multiple PDZ-domain containing scaffolding protein GRIP/ABP 88. Interestingly, GRIP can also bind to KIF1A through GRIP interacting protein liprin-α 91, suggesting that different microtubule-dependent motors are involved in AMPA receptor cargo transport. Moreover, GRIP contains up to seven PDZ domains and might transport many other interacting proteins, such as EphB receptors, their ephrin ligands and transmembrane protein Fraser syndrome 1 (FRAS1). Recently, GRIP has been implicated in dendrite morphogenesis by serving as an adaptor protein for EphB receptors 92. Synaptic scaffolding protein GKAP interacts with dynein light chain 93 and PSD-95 and related PDZ domain containing scaffolding proteins associate with KIF1Bα, however their function as a motor-cargo adaptors has not been shown directly 94 (Table 1). However, PDZ domain containing protein SAP102 has been implicated in NMDA receptor trafficking 95.

Motor proteins also make use of small Rab GTPases, often with the help of Rab effector molecules, to attach to their cargo. Several secretory and endosomal Rab proteins have been shown to bind to specific motors 96979899100101102103 (Table 1). Recently, it was shown that Rab3 effector DENN/MADD is an important linker between KIF1Bβ/KIF1A motors and Rab3-containing synaptic vesicle precursors 104. Surprisingly, the number of synaptic vesicles is not reduced in quadruple Rab3A-D knockout mice 105, suggesting that redundant synaptic vesicle trafficking mechanisms are present. Recently, Rab6 was shown to regulate transport and targeting of secretory vesicles 106. Rab6 binds the dynein/dynactin motor complex through its effector protein Bicaudal-D (BICD) and regulates microtubule-minus end directed transport 92107108. Moreover, recycling endosome GTPase Rab11 was shown to interact with myosin Vb and regulate AMPA receptor trafficking 45109. Myosin Vb binds to Rab11 through its effector Rab11-family interacting protein 2 (Rab11-FIP2) allowing the transport of AMPAR containing recycling endosomes into the dendritic spine 45.

One way to control intracellular trafficking is to regulate motor-cargo interactions by responding to changes in local ionic concentrations. Activation of NMDA receptors causes a rapid influx of Ca²⁺ in dendritic spines 114. A recent study shows that Myosin Vb is a "Ca²⁺ sensor" for actin-based postsynaptic AMPA receptor trafficking 45. Increased Ca²⁺ levels lead to unfolding of Myosin Vb motors and allows for binding to Rab11-FIP2 adaptors on recycling endosomes 115116 (Figure 2B). The association of myosin Vb with Rab11-FIP2 recruits AMPA receptor containing recycling endosomes into spines. Thus, elevated Ca²⁺ levels in spines promote actin-based postsynaptic trafficking. On the other hand, Ca²⁺ influx reduces mitochondrial motility 117. Recent studies suggest that the mitochondrial Miro-Milton adaptor complex is essential for the Ca²⁺-dependent regulation of mitochondria trafficking 118. Elevated Ca²⁺ levels permit Miro to interact directly with the motor domain of KIF5 (Figure 2B). The interesting aspect of this model is that KIF5 remains associated with mitochondria regardless of whether they are moving or stationary. In the "moving" state, KIF5 is bound to mitochondria by binding to Milton, which in turn interacts with Miro on the mitochondrial surface. In the "stationary" state, in presence of high Ca²⁺ levels, the KIF5 motor domain interacts directly with Miro and prevents microtubule interactions. Both findings imply the existence of "Ca²⁺ sensors" that detect neuronal activity stimuli and convert Ca²⁺ influx into mechanisms regulating cargo trafficking.

Ca²⁺ influx also triggers synaptic signaling cascades 119120. A key player in this process is calcium/calmodulin-dependent protein kinase II (CaMKII), a calcium-activated serine/threonine kinase that is present at synaptic terminals. Recent findings link CaMKII activity with regulated motor-cargo binding. For instance, CaMKII has been shown to phosphorylate KIF17, which induces the dissociation of the Mint1 scaffolding protein complex and release of NMDA receptor-containing cargo near the postsynaptic membrane 90 (Figure 2C). In this way, regulated CaMKII activity provides an attractive mechanism for targeting NMDA receptor complexes to active synapses where the kinase is switched "on". CaMKII has also been shown to modulate myosin V - cargo binding in Xenopus melanophores 121. It will be interesting to test whether myosin V is important for NMDA receptor-containing vesicle transport. Another recent study shows that CaMKII controls the trafficking of leukocyte common antigen-related (LAR) family of receptor protein tyrosine phosphatases (Figure 2C) 122. LAR associates with KIF1A via the adaptor protein liprin-α1 91123124. Interestingly, liprin-α1 is degraded in response to CaMKII phosphorylation and additionally regulated by the ubiquitin-proteasome system 122. At synapses where CAMKII is active, LAR-liprin-α1 containing vesicles would be "unloaded" due to CaMKII-mediated degradation of liprin-α1. An attractive idea is that the regulated degradation of cargo-adaptors or motor proteins is a general molecular mechanism for directed motor trafficking.

Rab proteins are molecular switches which cycle between a GDP-bound and a GTP-bound state and represent the inactive and active states, respectively 125. Activation of Rab GTPases is mediated by predominantly membrane-associated Rab guanine nucleotide exchange factor GEFs, inactivation of Rabs is a consequence of their intrinsic capacity to hydrolyse GTP. However, this activity is low and requires stimulation by Rab GTPase-activating proteins (GAPs). An additional level of regulation exists through guanine nucleotide dissociation inhibitors (GDIs), which prevent activation of GTPases by blocking their interaction with GEFs and effector proteins. Recent studies show that motor-cargo interactions are controlled by regulated Rab GTPase activity.

Collapsin response mediator protein-2 (CRMP-2) is crucial for axon formation in hippocampal neurons and functions as a motor-adaptor protein to link KIF5 to neurotrophin Trk receptors 126. CRMP-2 forms a complex with Rab27 and its effector Slp1, associates with the cytoplasmic region of TrkB and regulates anterograde transport of TrkB 127. Interestingly, the direct binding of TrkB to Slp1 is dependent on the GTPase activity of Rab27. The binding between Slp1 and GTP-bound Rab27 might be regulated by a Rab27 GAP mediated block of Rab inactivation (Figure 2D). On the other hand it is possible that Rab27 specific GDIs and GEFs regulate the recruitment to TrkB-containing vesicles. Another study shows that the nucleotide-bound Rab3 regulates trafficking of KIF1Bβ/KIF1A-positive synaptic vesicle precursors through the interaction with its effector protein DENN/MADD 104. GTP-Rab3 is more effectively transported than GDP-Rab3 in axons and Rab3 activity can regulate the interaction between KIF1Bβ/KIF1A-DENN/MADD and Rab3-containing synaptic vesicle precursors (Figure 2D). It will be interesting to find signaling pathways leading to the regulation of Rab3 GTPase activity. Interestingly, DENN/MADD has been shown to function as a GEF for Rab3 128. Identifying GEFs and GAPs for specific Rabs will be a major future challenge to understand how Rab GTPases regulate motor-cargo binding and transport specificity.