The human genome has more than 40 kinesin genes whose protein products organize intracellular traffic along microtubules. Research during the past two years has begun to elucidate the cargoes carried by kinesins and the nature of the kinesin-cargo linkage. Modular protein-protein interactions connect kinesins to diverse cellular molecules, which, apart from their other functions, serve as kinesin-cargo linkers. Many of these newly identified linkers are scaffolds for signaling pathways, and mounting evidence now indicates that kinesins transport pre-assembled signaling modules as vesicular cargo. These findings bring together two fields, signal transduction and molecular motors, and lead to a deeper understanding of the interplay between trafficking, localization and intercellular communication.
Kinesins constitute a family of molecular motors that contain a signature∼ 340 residue motor domain that transduces ATP hydrolysis into a directed walk along a microtubule (Vale and Milligan, 2000). Outside the conserved motor domain the primary structures of kinesins diverge, and it is this part – the tail – that binds cargo and regulates motor activity. Slowly but surely, progress is being made toward meeting the field's principal challenge: identifying the cargoes and defining how they are linked to, and released by, particular kinesins at the right time and place.
For many years it was presumed there were `kinesin receptors', that is, molecules whose sole purpose is to link kinesins to cargo. This presumption is not supported by the recent work, which instead indicates that kinesin-cargo linkers (Bowman et al., 2000; Kamal et al., 2000; Lee et al., 2002; Nakagawa et al., 2000; Setou et al., 2000; Setou et al., 2002; Verhey et al., 2001) are `familiar faces' that have other functions. Most are adaptors or scaffolds. Through their multiple protein binding sites, they organize molecular assemblies that constitute the cargo and are themselves part of the cargo (Fig. 1).
One of the first proteins to illustrate this new concept was the adaptor protein AP-1. Through its interactions with clathrin and specific receptor-ligand complexes (Pearse, 1988), AP-1 coordinates the formation of specific vesicle populations at the Golgi apparatus and the plasma membrane (reviewed in Kirchhausen, 2002). The new idea contributed by kinesin research (Nakagawa et al., 2000) is that a site on AP-1 also interacts with the tail of at least one kinesin– KIF13A. This interaction explains how a subset of Golgi-derived vesicles (those containing the mannose-6-phosphate receptor and its ligands) is transported along microtubules to the pre-lysosomal compartment (Nakagawa et al., 2000). This is an appealing idea: the same protein – AP1 – that sorts membrane proteins into specific cargo vesicles also interacts with the motor that transports these vesicles to their final destination.
The idea that kinesins are linked to adaptors that possess binding sites for multiple proteins suggests we may have underestimated the role of microtubule-based transport in cell biology. Given the large number of scaffold proteins and kinesins encoded by the human genome, the range of proteins that in principle could be part of one cargo or another is vast. As the kinesin-cargo linker molecules and the proteins they scaffold come to light, our notion of what constitutes `cargo' for intracellular transport along microtubules will inevitably broaden. For example, recent studies indicate that pre-assembled signal transduction cascades, or `transducisomes' (Tsunoda et al., 1997), are cargo for kinesins. Below, I discuss this novel role for kinesins at the interface of signaling and transport.
Trafficking of scaffolds for signaling pathways
Most signaling pathways employ scaffold proteins of one kind or another, which helps explain a fundamental tenet of intercellular signaling: from a common collection of enzymes, individual cells assemble parallel signaling pathways that respond to different stimuli and produce distinct physiological responses (Smith and Scott, 2002). A classic example is the scaffolding of MAP kinase (MAPK) pathways (van Drogen and Peter, 2002), which first came to light in studies of the pheromone mating response in Saccharomyces cerevisiae (Choi et al., 1994). The MAP kinase cascade (i.e. MAPK kinase kinase, MAPK kinase, and MAPK) controlling this response requires a scaffold protein, Ste5, which holds the three kinases in juxtaposition (Elion, 2001). This complex ensures that these kinases phosphorylate only their specific targets and not other potential targets in the cell. MAP kinase cascades that regulate other physiological processes are similarly organized on their own scaffolds. In this way, multiple MAP kinase signaling pathways, each coupled to a different receptor, can be isolated from each other, while employing some of the same components.
Scaffold molecules are particularly important for the organization of PDZ-based signaling cascades. Many of these reside at specific subcellular sites in the nervous system, where the need for a high degree of localization, for example, at synaptic junctions, is obvious. Equally important is the need to localize signaling cascades in epithelia. Many intercellular signaling pathways that govern cell fate specification during embryonic development occur in the context of epithelia (Fig. 2). Like neurons, epithelial cells are asymmetric: the basolateral and apical domains encounter distinct extracellular environments. Consequently, communication from an inducing cell to a neighboring cell will depend on localization of the relevant signaling molecules.
The importance of trafficking and localizing signaling modules seems obvious in these contexts and is supported by a great deal of circumstantial evidence. However, only a few studies have actually provided direct evidence that proper trafficking of signaling molecules to a highly localized subcellular destination is essential for intercellular signaling and global expression of signaling molecules is inadequate. Among the clearest cases is the seminal study that identified a complex of scaffold proteins, known as the LIN complex, in the context of vulva development in Caenorhabditis elegans (Kaech et al., 1998). In C. elegans, a canonical Ras MAP kinase signaling pathway specifies an epithelial precursor cell to become a vulva cell. This pathway is activated when an EGF receptor tyrosine kinase, LET27, in the precursor cell binds EGF released by an `anchor cell'. Because the anchor cell resides in the stroma (Fig. 2), the EGF it releases cannot penetrate the intercellular junctions connecting the precursor cells. The products of the LIN2, LIN7 and LIN10 genes prevent differentiation of the precursors by interfering with their ability to establish a localized cluster of LET27 at the basolateral aspect of the junctional complex. These three proteins form an evolutionarily conserved PDZ-based scaffold complex that interacts with LET27, as well as a number of different receptors in various organisms and cell types (Bredt, 1998). In worms, null mutations in the LIN2, LIN7 or LIN10 genes lead to the same phenotype: LET27 is still present in the precursor cells, and even exists in the basolateral compartment, but it is diffusely distributed and not concentrated at the cell junction. Second allele complementation experiments elegantly prove that specific interactions between the LIN scaffold complex and LET27 are required for proper localization of the receptor LET27 and operation of the pathway (Kaech et al., 1998). These studies provide direct experimental evidence that expression and global distribution of the relevant signaling molecules in a cell are not sufficient for signaling to occur normally. These molecules need to be properly trafficked and localized.
In summary, it is clear that signaling cascades must be trafficked and localized, and that these activities involve scaffold proteins. But how is the localization of signaling modules achieved? And exactly where in the cell, and by what mechanisms, do signaling molecules load onto scaffolds? Until recently, it was widely presumed that scaffold proteins and their signaling molecule partners arrive at their destinations, where they assemble locally, by diffusion. Recent research on kinesin-cargo interactions instead suggests that signaling molecules are loaded onto their scaffolds away from their final destinations and that active transport along microtubules delivers these pre-assembled signaling modules to particular destinations. I will review how these new ideas developed from efforts in the kinesin field to define the nature of the motor-cargo linkage and the identity of the cargoes carried by kinesins, discussing this recent work chronologically because of the compelling manner in which the experiments and ideas progressed.
KIF 17 traffics glutamate receptors in the brain
The first definitive evidence that signaling molecules are cargo for kinesins developed from a yeast two-hybrid screen for partners of the mouse neuronal kinesin KIF17 (Setou et al., 2000). The KIF17 C-terminus has a PDZ-binding motif that binds the first PDZ domain of mouse brain LIN-10, that is, the homologue of C. elegans LIN-10 mentioned above. Prior to the discovery that mouse brain LIN-10 interacts with KIF17 (Setou et al., 2000), several laboratories had established that the complex of LIN-10/Mint1, LIN-2/CASK and LIN-7/Velis is evolutionarily conserved both in its structure and in its association with transmembrane receptors that are localized to discrete subcellular sites (Bredt, 1998); for example, in C. elegans, the LIN complex is required for targeting EGF receptors to cell junctions in vulval precursor cells (Kaech et al., 1998) and for localizing glutamate receptors to the postsynaptic density in neurons (Rongo et al., 1998). Although other studies also raised the possibility that the LIN complex localizes glutamate receptors to postsynaptic densities in the mammalian brain (Jo et al., 1999), the mechanism(s) underlying this and other LIN-associated localization events have not been resolved.
In light of these previous studies, the discovery that LIN-10 interacts with KIF17 immediately raised the possibility that postsynaptic clusters of glutamate receptors are established, at least in part, through KIF17-driven transport (Setou et al., 2000) (Fig. 3). In this context, the LIN complex serves as a kinesin-cargo linker: it connects KIF17 to the transmembrane protein NR2B (a subunit of the NMDA-sensitive glutamate receptor), which itself provides the connection to a vesicle (Figs 1 and 3). In vitro vesicle motility experiments support this model (Setou et al., 2000): dominant inhibitory KIF17 constructs inhibit nucleotide-dependent microtubule binding of NR2B-containing vesicles isolated from the brain; and wild-type KIF17, but not mutants lacking the C-terminal residues that bind LIN-10, promotes the motility of isolated NR2B-containing vesicles along microtubules.
It is unclear whether the LIN complex also scaffolds signaling molecules downstream of the receptors it carries, but other multi-PDZ domain proteins do function in this manner (Harris and Lim, 2001; Sheng and Sala, 2001). For example, InaD has five PDZ domains that each hold a component of the phototransduction cascade in Drosophila photoreceptors (Tsunoda et al., 1997). Generation of a normal light response depends on this architecture. Mutations within single PDZ domains produce specific defects in the light response, each associated with mislocalization, to the cytoplasm or plasma membrane, of the PDZ partner. InaD thus localizes, and is responsible for the targeting of, a highly organized transducisome (Tsunoda et al., 1997).
Conventional kinesin (kinesin I)
The idea that the LIN scaffold complex is also a KIF17-cargo linker (Setou et al., 2000) provided the first indication of a relationship between kinesins and the localization/trafficking of signal transduction scaffolds. Subsequent work on kinesin I generalized and extended this concept (Fig. 3). Kinesin I was originally discovered in the context of vesicle transport in axons (Schnapp et al., 1985; Vale et al., 1985), but its in vivo function and cargo were uncertain for many years. It now appears that this kinesin carries multiple signaling modules as vesicular cargo (Fig. 1).
Kinesin I structure, motility and regulation
Native kinesin I is a tetramer consisting of a kinesin heavy chain (KHC) dimer and two kinesin light chains (KLCs) (Fig. 4). The KLCs are non-motor polypeptides that associate with the KHC dimer exclusively, that is, they are not found at appreciable levels, apart from the KHC dimer, and, as a general rule, the KHC dimer is always associated with KLCs (Hackney et al., 1991). For many years the functions of the KLCs were enigmatic, but we now know they participate in two activities that presumably are coordinated: linking kinesin I to cargo and regulating kinesin I motor activity.
As the vesicle-microtubule interface can only accommodate one or two motor molecules, kinesin I (Block et al., 1990; Howard et al., 1989) and other kinesins that carry small cargo (Tomishige et al., 2002) must be processive, that is, they must walk along the microtubule without letting go. If this were not the case, continuous transport of cargoes over long distances would not be possible because the kinesin-cargo complex would diffuse away from the microtubule. Although processivity would seem to imply that the bulk of kinesin I should be bound tightly to microtubules, most kinesin I is actually soluble, unbound to either microtubules or cargo. Moreover, the microtubule-stimulated ATPase rate of isolated kinesin I is too low (Hackney, 1995) to account for the ∼600 nm sec–1 rate of kinesin-driven movement, given that one ATP hydrolytic event powers each 8 nm step along the microtubule (Schnitzer and Block, 1997; Svoboda et al., 1993). What reconciles these apparent inconsistencies is that kinesin I motor activity is turned off in the absence of cargo. If it were not, the motor population would walk to the end of the tracks and accumulate at the cell periphery. By negatively regulating motor activity, the cell maintains the concentration of a processive motor such as kinesin I at a uniform level throughout the cell and in large excess of cargo. An appealing hypothesis is that a cargo controls its own destiny by binding to and activating motors on demand (Verhey et al., 1998).
Kinesin I appears to be turned off in the absence of cargo by a self-inhibition mechanism that depends on its tail and involves folding of the KHC dimer (reviewed in Verhey and Rapoport, 2001; Woehlke and Schliwa, 2000) (Figs 4 and 5). The KHC dimer contains coiled coils interrupted by short unstructured regions that act as hinges (Fig. 4). Isolated native kinesin I is primarily in a folded conformation (Fig. 5) that prevents the motor domains from engaging the microtubule, and thereby inhibits motility and ATP hydrolysis. The principal evidence supporting this model comes from mutational analyses, which indicate that hinge 2 (Friedman and Vale, 1999; Seiler et al., 2000) and C-terminal residues of KHC (Hackney and Stock, 2000; Seiler et al., 2000; Stock et al., 1999; Verhey et al., 1998) are essential for inhibition. The C-terminal residues, which carry excess positive charge, might interact with a region of excess negative charge in the neck coiled coil (Stock et al., 1999) to stablilize the folded conformation.
Although such self-inhibition by folding is an integral property of the KHC dimer, the KLCs are needed for full repression of KHC in vivo (Verhey et al., 1998): in the absence of KLCs, epitope-tagged KHCs colocalize with microtubules and accumulate at the ends of cell processes, that is, at the minus ends of microtubules. In vitro experiments also support the idea that the KLCs contribute to inhibition (Friedman and Vale, 1999). Although the exact mechanism by which KLCs inhibit kinesin I motor activity is unclear, the KLC structure-function relationships provide some clues (Verhey et al., 1998). KLCs consist of at least two structurally distinct regions (Fig. 4): an N-terminal region of heptad repeats and a C-terminal region that has six tetratrico peptide repeats (TPRs) – protein-binding modules present in diverse proteins (Blatch and Lassle, 1999). The heptad repeats alone are sufficient for binding KHC (Fig. 4) and inhibiting the ability of KHC to engage microtubules. By contrast, the KLC TPRs are required for neither of these activities. In view of the idea that at some point KHC motor activity must be activated in relation to cargo binding, it was appealing to consider that the partner(s) of the KLC TPRs might be involved in cargo binding or motor activation (Verhey et al., 1998). Thus, identifying the partners of the KLC TPRs became a priority.
Kinesin I carries a MAP kinase cascade
We now know that the KLC TPRs are kinesin I cargo-binding domains. Initially this came to light following the discovery that the TPRs interact with a family of MAP kinase scaffold proteins – the JIPs (JNK-interacting proteins, known also as JSAPs) (Bowman et al., 2000; Verhey et al., 2001). The three JIP isoforms in mammals are scaffolds for the MAP kinase cascade that activates Jun N-terminal kinase (JNK) (Davis, 2000). In common with InaD and Ste5, JIPs juxtapose a cascade of kinases, ultimately enabling the efficient and specific phosphorylation of JNK at two sites by a MAPKK (Davis, 2000). Activated JNK phosphorylates downstream targets, including the transcription factor Jun, and regulates various physiological activities (Weston and Davis, 2002), including apoptosis in the brain (Morishima et al., 2001). Among the many different transmembrane receptors implicated in JNK signaling is ApoER2, the Reelin receptor (Stockinger et al., 2000). The Reelin signaling pathway has a very important role in neurogenesis (Rice and Curran, 2001), which is consistent with the enrichment of kinesin I and JIPs in the brain.
The KLC TPRs recognize a motif found at the extreme C-termini of JIP 1 and JIP2 (Fig. 3) (Verhey et al., 2001). TPR domains in other proteins likewise recognize the C-termini of their partners (Gatto et al., 2000; Scheufler et al., 2000; Terlecky et al., 1995). This establishes a striking parallel between the kinesin-I–JIP linkage and the KIF17-LIN scaffold linkage (Fig. 3) with respect to how kinesin-cargo specificity is governed. Both interactions involve binding modules, PDZ domains or TPRs, that recognize motifs in the C-termini of their partners. In the case of kinesin I, the module is in the kinesin and the motif in the linker, whereas in the case of KIF17, the module is in the linker and the motif in the kinesin. But the interactions are comparable fundamentally.
The significance of the JIP-KLC TPR interaction came to light in three complementary studies (Bowman et al., 2000; Byrd et al., 2001; Verhey et al., 2001). First, Bowman and co-workers identified Sunday Driver (SYD) (Bowman et al., 2000), the Drosophila homologue of mammalian JIP3, in a mutant screen for genes that produce in larvae a behavioral phenotype called tail flipping, which is characteristic of kinesin I KHC and KLC null nutants (Hurd and Saxton, 1996; Saxton et al., 1991). This phenotype is associated with axonal jams – accumulations of membrane organelles in peripheral nerves (Hurd and Saxton, 1996) – and arises as the level of maternal kinesin I decline during development, which leads to death. Although the exact relationship between axonal jams and a specific kinesin I transport function is unclear, the similar phenotypes of KLC, KHC and SYD mutants indicate that the products of the three genes are involved in the same process. The additional finding that SYD co-precipitates with kinesin I in vivo and interacts directly with KLC TPRs suggested SYD functions by interacting directly with kinesin I. But these studies left open the question of whether SYD/JIP3 and its associated signaling molecules are kinesin I regulators, cargoes or both.
Simultaneously, studies by Verhey et al. (Verhey et al., 2001) provided evidence that JIPs are cargoes of kinesin I in mammals. Earlier work had indicated that JIPs are highly enriched at the tips of neuronal processes (Kelkar et al., 2000; Meyer et al., 1999), that is, at the minus ends of microtubules, suggesting that they are transported there by kinesin I. The fact that dominant inhibitory KLC constructs (heptad repeats without TPRs or TPRs without heptad repeats) block neurite tip localization of JIPs supports the hypothesis that JIPs are kinesin I cargo (Verhey et al., 2001). In addition, the effects of mutations in JIP1 or JIP2 C-terminal residues on KLC TPR binding and on neurite tip localization correlate (Verhey et al., 2001). Thus, the steady-state distribution of JIP1 and JIP2 in neuronal cell lines depends on the interaction between the conserved C-terminal residues of JIP1 and JIP2 and the KLC TPRs of kinesin I.
These studies generalize the model of kinesin-cargo architecture established initially in the context of KIF17 (Fig. 3) and also contribute a new idea: the JIP scaffold is pre-loaded with its kinase cascade prior to transport. This idea was brought to light by the finding that not only JIP but the kinases scaffolded by JIPs, and the transmembrane receptor ApoER2, all co-precipitate from brain extracts with kinesin I when the latter is isolated either with an antibody or by nucleotide-dependent microtubule co-sedimentation (Verhey et al., 2001). Furthermore, kinesin I dominant inhibitory constructs that inhibit neurite tip localization of JIP likewise inhibit localization of the MAPKK scaffolded by JIP (Verhey et al., 2001). These findings fit nicely with the notion of a transducisome (Tsunoda et al., 1997) and raise the possibility that transducisomes are also trafficking units. The findings also support the idea that signaling scaffolds, in addition to juxtaposing kinases in a cascade, carry information about the trafficking and localization of the cascade. This model differs from the conventional view that signaling molecules assemble on scaffolds at their final destination.
Does signaling through the cascade feedback to regulate the motor? One appealing proposal (Verhey and Rapoport, 2001) is that activation of the Reelin pathway by ApoER2 causes kinesin I to dissociate from the JIP scaffold upon fusion of the vesicle with the nerve terminal membrane. Although this specific hypothesis has not been tested, a third line of investigation (Byrd et al., 2001) establishing an interaction between JIP scaffolding proteins and kinesin I indicates that kinases scaffolded by JIP regulate intracellular vesicle traffic, including cargo transport by kinesin I. These studies identified unc-16, which encodes the JIP 3/SYD homologue, in a genetic screen for molecules that organize presynaptic terminals in the C. elegans nervous system. In C. elegans, as in mammals and flies, JIP 3 scaffolds JNK and its upstream kinases and is dependent on kinesin I for its normal distribution in neurons (Byrd et al., 2001). Partial loss-of-function alleles of kinesin I, unc-16 or JNK and its upstream kinases produce the same phenotype– mislocalization of synaptobrevin (Byrd et al., 2001) – and together enhance this phenotype; hence these genes must function in the same process. Signaling via kinases carried by kinesin I thus does regulate vesicle transport, but the nature of this regulation is unclear. Whether kinesin I itself is regulated directly by the kinases it carries is yet to be addressed.
One kinesin – multiple cargo-linkers
A key question is whether each member of the kinesin family is committed to carrying one or multiple cargoes. Investigations of kinesin I support the latter idea. For example, the KLC TPRs interact not only with JIPs but with amyloid precursor protein (APP) (Kamal et al., 2000). APP is a transmembrane protein that in vivo is subjected to proteolytic cleavages, including a final, intramembrane, presenilin-dependent cleavage (Selkoe, 1998). This cleavage produces the extracellular amyloidβ -peptide of Alzheimer's disease and releases an intracellular tail fragment of unknown function. Despite intensive research, the normal cellular function of APP is not known, but the proteolytic processing of APP is strikingly similar to that of Notch (Artavanis-Tsakonas et al., 1999), the transmembrane receptor for a signaling pathway involved in cell fate specification. Cleavage of Notch produces a cytoplasmic fragment that enters the nucleus and regulates transcription. Other signaling molecules, for example, E-cadherin (Marambaud et al., 2002) and the ErbB-4 receptor tyrosine kinase (Ni et al., 2001), are similarly processed. Recent studies extend the parallel between Notch and APP by demonstrating that the cytoplasmic tail of APP forms a multimeric complex with the nuclear adaptor protein Fe65 and the histone acetyltransferase, Tip60 (Cao and Sudhof, 2001).
Definitive evidence now demonstrates that APP is transported in axons by kinesin I (Gunawardena and Goldstein, 2001; Kamal et al., 2000). Earlier studies (reviewed in Selkoe, 1998) demonstrated that APP is synthesized in the endoplasmic reticulum, glycosylated in the Golgi apparatus, and packaged into vesicular structures that are transported down axons (Yamazaki et al., 1995). Transport is blocked by anti-sense oligos against kinesin I (Amaratunga et al., 1993; Ferreira et al., 1992). The new work establishes that the 47 C-terminal residues constituting the cytoplasmic domain of APP interact directly with the KLC TPRs (Kamal et al., 2000) and that kinesin I is associated with a defined class of axonally transported vesicles that contain APP and the transmembrane proteins involved in its proteolytic cleavage (Kamal et al., 2001). That the interaction between APP and KLC is required for APP transport is clear from several experiments. Sciatic nerves from mice lacking one of the three known KLC isoforms (KLC1, the neuronally enriched isoform) transport less APP than nerves from normal animals (Kamal et al., 2000) and, in Drosophila, deletion of the APP-like (appl) gene produces the axonal jam phenotype in larvae that characterizes khc and klc mutants (Gunawardena and Goldstein, 2001). Finally, the fraction of endogenous APP that has been phosphorylated at Thr 668 by the neuronal kinase CDK5 (Iijima et al., 2000) accumulates at the tips of neurites in mammalian neuron-like cells in culture (Ando et al., 1999), and the localization of this modified APP is abolished by overexpression of dominant inhibitory KLC constructs (e.g. the TPRs or heptad repeats of KLC) (Muresan et al., 2001). This finding, together with the observation that phosphorylation of APP Thr 668 by CDK5 promotes the interaction between the APP C-terminus and KLC TPRs in vitro (Muresan et al., 2001), supports the hypothesis that the transport of APP requires an interaction between the cytoplasmic domain of APP and the KLC TPRs. The idea that APP is a kinesin-I-cargo linker establishes a precedent for direct interactions between kinesins and transmembrane proteins and thus suggests that the generalized architecture in which kinesins are linked to transmembrane proteins indirectly through soluble scaffolds (see Fig. 1), although common, is not universal.
Still unclear is whether APP and the JIP scaffold are carried on the same cargo vesicles. Structural studies of the TPR motif indicate that three TPR repeats would be sufficient to bind a partner (Gatto et al., 2000; Lapouge et al., 2000; Scheufler et al., 2000); hence it is conceivable that APP and JIPs bind simultaneously to the six TPR repeats in a single KLC. Alternatively, a single kinesin I motor could carry JIPs and APP simultaneously by devoting one KLC to each partner. A related question is whether the JNK signaling cascade carried by kinesin I and APP are involved in the same signaling pathway.
The KLC TPRs do not provide the only cargo-binding site on kinesin I. Earlier studies focusing on Neurospora kinesin I, which lacks KLCs, identified a candidate cargo binding region on the KHC (Fig. 4) (Seiler et al., 2000). Seiler et al. searched for functional KHC domains, using KHC deletion mutant cDNAs expressed as transgenes and evaluating their ability to rescue the reduced growth rate of a KHC-deficient strain of Neurospora. They identified in the tail coiled coil a domain of 51 residues (Fig. 4) that is highly conserved among KHCs from different species. In Neurospora, this region is essential for localizing tagged KHC constructs to small vesicles destined for secretion at the hyphal tip (Seiler et al., 2000). Setou et al. subsequently used this region in a yeast two-hybrid screen of mouse brain cDNAs (Setou et al., 2002) to identify the glutamate receptor interacting protein 1 (GRIP1), a known multi-PDZ-domain protein at synaptic junctions (Dong et al., 1997; Srivastava et al., 1998), as the kinesin I partner. A region of GRIP1 between the sixth and seven PDZ domains binds to a region of KHC that overlaps the cargo-binding domain defined originally in Neurospora (Fig. 4).
GRIP1, and the related protein GRIP2, scaffold several neuronal signaling proteins (Wyszynski et al., 2002). In particular, the fifth GRIP1 PDZ domain interacts with the C-terminal sequence (-ESVKI) of subunits 2 and 3 (GluR2/3) of AMPA-sensitive glutamate receptors (Wyszynski et al., 1999), which mediate excitatory synaptic transmission in the brain. Indeed, GRIP was initially identified through this interaction, and a large fraction of GluR2 exists as a complex with GRIP (Wyszynski et al., 1999). Although its exact function is unclear, GRIP presumably contributes to the localization, at postsynaptic densities, of a large multi-protein complex that transduces presynaptic activity into postsynaptic responses.
GRIP1, GluR2 and kinesin I are colocalized within the dendrites and cell bodies of cultured hippocampal neurons, and co-immunoprecipitate from vesicle fractions (Setou et al., 2002). That the interaction between kinesin I and GRIP1 is functionally important for trafficking of AMPA-senstitive glutamate receptors is evident from dominant inhibitory experiments: overexpression of the KHC GRIP1-binding site reduces the amount of both GRIP1 and GluR2 in dendrites.
The work described above provides further support for a generalized model of kinesin-dependent trafficking of signaling modules that involves vesicles and soluble scaffolds. It also raises interesting questions regarding the logic of this trafficking. Using binding sites on KHC, kinesin I carries AMPA-sensitive glutamate receptor vesicles within dendrites, but using binding sites on KLCs it carries JIP or APP vesicles in axons. How does kinesin I, which is distributed uniformly in both axons and dendrites, transport some cargo vesicles to dendrites and others to axons? One interesting proposal is that occupation of a particular cargo-binding site directs the motor to distinct microtubules in axons and dendrites (Setou et al., 2002).
One cargo – multiple kinesins?
The interaction between GRIP and kinesin I is apparently not the sole means of delivering AMPA-sensitive glutamate receptors to postsynaptic densities. GRIP is also carried by KIF1A (Lee et al., 2002) indirectly through the neuronal scaffold protein Liprin-α. The sixth PDZ domain of GRIP interacts with the Liprin-α C-terminus (Wyszynski et al., 2002), and the Liprin-α N-terminal coiled coil interacts with the KIF1A tail between residues 650 and 1105 (Lee et al., 2002). This KIF1A region overlaps with a recently identified protein-binding domain termed the MAGUK-binding stalk domain or MBS (Asaba et al., 2003) (Fig. 4), which warrants an explanation before we return to the transport of AMPA-sensitive glutamate receptors by KIF1A.
In addition to their presence in non-motor proteins, MBS domains are found in all members of the KIF1 subfamily of kinesins (Asaba et al., 2003), next to their FHA domains (Fig. 4). The MBS domain was originally identified in GAKIN (Hanada et al., 2000) – the human homologue of KIF13b – through its interaction with the protein human disks large (hDlg) (Asaba et al., 2003). Dlg is a member of the membrane associated guanylate-kinase MAGUK family of scaffold proteins, which are composed of one or more PDZ domains, an SH3 domain and a guanylate kinase-like (GUK) domain that lacks enzymatic activity and instead functions as a protein-binding module (Anderson, 1996). In humans, the KIF13b MBS interacts with the GUK domain of Dlg (Asaba et al., 2003). This interaction is extremely interesting because Dlg and other MAGUK proteins localize signaling complexes at specialized membrane sites such as tight junctions and synaptic junctions (Muller et al., 1996). For example, Drosophila Dlg and its orthologues scaffold large protein complexes at pre- and postsynaptic membranes and at the basolateral membrane of epithelial cells (Lue et al., 1994). The scaffold protein LIN 2 (Fig. 2) is also a MAGUK protein, and as discussed above, in C. elegans it localizes an EGF receptor at tight junctions and glutamate receptors at synaptic junctions. The finding that MBS domains exist in all KIF1 kinesins, and the demonstration that the human KIF13b-MBS–Dlg interaction is needed to maintain Dlg localization in MDCK cells (Asaba et al., 2003), establishes a relationship between KIF1 kinesins and the MAGUKs. The finding that the Liprin-α-binding site overlaps the MBS in KIF1A suggests that the MBS also interacts with scaffold proteins that lack GUK domains. Thus, one can imagine that the MBS of KIF1A, like the KLC TPRs of kinesin I, interacts with more than one type of cargo linker.
The GRIP–Liprin-α interaction is clearly necessary for localization of AMPA-sensitive glutamate receptors at postsynaptic sites on the dendrites of cultured hippocampal neurons, as overexpressed GRIP or Liprin-α constructs lacking the interaction sites diminish the population of localized AMPA-sensitive glutamate receptors (Wyszynski et al., 2002). The evidence that KIF1A transports glutamate receptors on the Liprin-α–GRIP scaffold is less direct. Liprin-α accumulates in the cell bodies of hippocampal neurons that overexpress a KIF1A construct lacking the motor domain (Lee et al., 2002). One presumes localization of AMPA-sensitive glutamate receptors would be diminished, but this has not been shown directly. Nevertheless, these studies again provide evidence for a signaling module carried as vesicular cargo by a kinesin. Because GRIP and Liprin-α have multiple binding sites for other proteins, this scaffold complex could in principle hold components of a signaling pathway downstream of the glutamate receptor or other receptors that interact with GRIP (reviewed in Wyszynski et al., 2002).
The implication that GRIP is carried by both KIF1A and kinesin I raises several questions. Do KIF1A and Kinesin I mediate parallel, functionally redundant transport pathways for the AMPA receptor? Or is AMPA receptor delivery achieved in two consecutive transport steps – might, for example, one motor transport the cargo from the Golgi apparatus to the primary dendrite, but the other mediate localized trafficking within the dendritic tree? Assuming GRIP is carried by KIF1A within dendrites, how does this one, ubiquitously distributed (Lee et al., 2002), motor also deliver synaptic vesicle precursors within axons (Hall and Hedgecock, 1991; Yonekawa et al., 1998)?
Summary and perspective
The recent progress on kinesin-cargo interactions provides a basis for understanding how kinesins interface with other aspects of cell biology, particularly protein trafficking and signal transduction. At these interfaces, a number of general principles have emerged. It now appears that many cellular molecules have adapted through evolution to serve as kinesin-cargo linkers. In several cases, modular protein-protein interactions govern these interactions. With one exception (APP), identified kinesin-cargo linkers are soluble scaffold proteins whose multiple binding sites organize functionally related proteins into a single cargo. These properties significantly broaden our expectations about the kinds of cargo carried along microtubules by kinesin motors, and thereby deepen our understanding of biological processes such as intercellular signaling. For example, the new work indicates that scaffolds for signaling pathways not only direct the flow of information in the cascade, but control trafficking and localization of the cascade through their interactions with kinesin tails. The findings that a single kinesin (e.g. kinesin I) has multiple cargo-binding sites and can carry diverse cargo, together with the idea that one molecule (e.g. AMPA receptor) can be cargo for different kinesins (KIF1A and kinesin I), suggest an unanticipated degree of plasticity to protein trafficking by motor proteins. The future is now rich with questions, whose resolution is likely to change our ideas about intracellular trafficking fundamentally. Chief among these is the question of exactly how and where cargoes are assembled, and the regulatory principles by which cargoes activate, and let go of, particular kinesins at the right time and place.
I am indebted to the members of my laboratory, particularly Virgil Muresan (now at Case Western) for many helpful discussions. Research in my laboratory is supported by grants from the NIH and a grant from the Human Frontiers Program.
- © The Company of Biologists Limited 2003