Myosin VI: two distinct roles in endocytosis

Endocytosis is an essential function in all cells and is required for nutrient uptake, receptor internalization and synaptic transmission. Immediately under the plasma membrane where endocytosis occurs is a cytoskeletal layer made up primarily of F-actin and associated proteins. You might predict that actin plays a role in endocytosis; however, the role of actin in this process has been controversial (reviewed by Qualmann et al., 2000), given evidence both for and against its importance, depending primarily on the cell type. Only in lower eukaryotes, such as yeast, have genetic and biochemical studies tightly linked the actin cytoskeleton to the endocytic process(reviewed by Qualmann and Kessels,2002). Recent studies on the unconventional myosin, myosin VI, an actin-based molecular motor expressed only in higher eukaryotes, now provide compelling evidence that this myosin, and therefore actin, is involved in two distinct steps in endocytosis. Here I highlight these findings and discuss mechanisms by which a myosin might be recruited to actin filaments to facilitate endocytic traffic in mammalian cells.

Myosins are a large family of structurally diverse molecular motors. All myosins share a conserved domain that binds to F-actin microfilaments and hydrolyzes ATP to produce movement along the filament. At least 18 distinct myosin classes have been identified to date(Berg et al., 2001), each having a unique C-terminus that presumably targets it to distinct cargo and subcellular regions.

Myosin function can be inferred in part from the direction of motor movement on actin filaments. At the cell cortex, where endocytosis occurs,polarized actin filaments are anchored with their barbed ends at the plasma membrane and the pointed ends facing inwards. Conventional myosins as well as unconventional myosins, such as myosins I and V, travel towards the barbed end of the actin filament. Thus, most myosins function in exocytosis or outward movement of organelles and not endocytosis. Only one newly identified unconventional myosin, myosin VI, has been shown to travel along actin filaments towards the pointed end (Wells et al., 1999). Its role as the only identified pointed-end directed molecular motor has been reviewed elsewhere(Cramer, 2000). Below I focus on its potential roles in endocytosis.

In polarized cells, myosin VI is associated with endocytic domains. In kidney proximal tubule cells and intestinal enterocytes, it is enriched at specialized clathrin-coated invaginations at the base of the brush-border microvilli (Biemesderfer et al.,2002; Buss et al.,2001; Heintzelman et al.,1994) (Fig. 1). These clathrin-rich invaginations are encased in the actin-rich terminal web,and within this region myosin VI overlaps with both clathrin and the clathrin-adapter protein AP-2 (Biemesderfer et al., 2002; Buss et al.,2001). In a very different polarized cell type, the hair cells of the inner ear, myosin VI is enriched in the pericuticular necklace, a vesicle rich region that is the primary site of endocytosis in this cell type(Hasson et al., 1997). Taken together these very divergent localization studies implicate myosin VI in endocytic processes.

Three distinct stages of endocytosis in polarized cells might require an actin-based motor like myosin VI (Fig. 2). The first is clustering of ligand-bound receptors into clathrin-coated pits (Fig. 2A). In the kidney, for example, the brush border microvilli are enriched in endocytic receptors. These receptors are involved in the uptake of amino acids, vitamins and other components from the filtered urine. Although these receptors are present along the length of the microvillus, they are also concentrated in the clathrin-rich apical invaginations between the microvilli(reviewed by Christensen et al.,1998). Pulse-chase studies following uptake of cationic ferritin in MDCK cells revealed that ferritin associates first with microvilli and later with apical intermicrovillar invaginations and clathrin-coated pits,suggesting that ligand-bound receptors are actively transported to the base of microvilli (Gottlieb et al.,1993). Microvilli are composed primarily of bundled actin filaments polarized with their barbed ends towards the tips of the microvilli(Mooseker et al., 1982). Regulated movement of ligand-bound receptors down the microvilli would require a pointed-end-directed myosin motor to bind and transport components directly down the microvillus. Because myosin VI is present at low levels along microvilli (Biemesderfer et al.,2002) as well as at their base, it may serve this role in directed receptor transport.

A second potential role for myosin VI is in the formation of clathrin-coated vesicles (Fig. 2B). Studies in proximal tubule cells following the scavenger receptor megalin after ligand binding revealed that the ligand is first concentrated in the clathrin-rich invaginations before traversing through an endosomal compartment and reaching the lysosome(Birn et al., 1997; Christensen and Nielsen,1991). The intermicrovillar invaginations are enmeshed in a dense actin cytoskeleton (the terminal web). This actin may act as a barrier to vesicle formation or as an active player in the vesicle formation process. Indeed, four distinct models for a positive role for actin in clathrin-coated vesicle formation have been proposed(Qualmann et al., 2000) (see below). Because myosin VI is concentrated in the clathrin-rich invaginations,it might function in any or all of these actin-dependent processes,facilitating the creation of clathrin-coated vesicles.

Finally, a third potential role for myosin VI is at a later step of endocytosis; the transport of uncoated vesicles through the actin-rich terminal web towards the early endosome(Aschenbrenner et al., 2003)(Fig. 2C). During the process of receptor-mediated endocytosis in polarized cells as well as cells that have a dense cortical actin network, a mechanism must be in place to facilitate movement of the uncoated vesicle from the peripheral region of the cell,through the actin meshwork, towards the more central early endosomes and microtubule networks for further transport. As a pointed-end directed motor present in the terminal web and peripheral actin networks, myosin VI is a good candidate for a vesicle motor at the heart of such a mechanism.

Of these three potential roles, studies in cultured cell models have provided strong evidence for two - formation of clathrin-coated vesicles(Fig. 2B) and transport of uncoated vesicles (Fig. 2C). The evidence implicating myosin VI in these two steps is based on studies of splicing variants that allow myosin VI to target to two distinct endocytic compartments (Aschenbrenner et al.,2003; Buss et al.,2001).

Myosin VI has four functional domains(Fig. 3). The conserved motor domain of myosin VI is at the N-terminus of the protein, and is followed by a single IQ motif, which serves as a light-chain-binding site for the calcium-binding protein calmodulin (Hasson and Mooseker, 1994). The motor-IQ domain is sufficient for pointed-end directed movement (Homma et al., 2001; Wells et al.,1999). Following the motor is the tail domain, which is made up of two regions: a 200-residue coiled-coil region and a C-terminal globular region(Hasson and Mooseker, 1994). The coiled-coil region mediates dimerization(De La Cruz et al., 2001). The globular region contains no recognizable motifs but is highly conserved across species.

The tail of class VI myosins is alternatively spliced in both insects and vertebrates (Breckler et al.,2000; Buss et al.,2001; Buss et al.,1998; Kellerman and Miller,1992), leading to insertions between the coiled-coil domain and the globular domain, as well as additional residues at the C-terminus. Although the significance of the alternative splicing has not been fully elucidated, in vertebrates alternative splicing generates two predominant forms of myosin VI, a longer and a shorter form, which differ by a 23-residue insert between the coiled-coil and the globular region(Fig. 3). Both the long and short forms of myosin VI have been implicated in endocytosis.

The longer splice form of myosin VI has been directly implicated in the formation of clathrin-coated vesicles during endocytosis(Buss et al., 2001)(Fig. 3). This form of myosin VI colocalizes with clathrin-coated pits when expressed in cultured cells, and PCR analysis suggests that it is the major form expressed in rat kidney and polarized CaCo-2 intestinal epithelial cells(Buss et al., 2001). Analysis of myosin VI fragments revealed that the tail domain alone does not target myosin VI to clathrin-coated pits, but the tail domain plus the splice site insert is sufficient for coated-pit targeting. Cultured NRK cells expressing a myosin VI fragment containing the tail domain with the splice insert exhibit a dramatic reduction (∼70%) in transferrin endocytosis(Buss et al., 2001). Therefore, alternative splicing of myosin VI generates a form of myosin VI unique to polarized cell types that can be recruited to function in the formation of clathrin-coated vesicles.

Epithelial cells cultured under non-polarizing conditions express the shorter version of myosin VI (Fig. 3) (Buss et al.,2001; Hasson and Mooseker,1994). Although these cells lack microvilli, they do exhibit dense cortical actin networks. In these cells, myosin VI is found on peripheral vesicles that pulse-chase analysis confirms are recently uncoated endocytic vesicles (Aschenbrenner et al.,2003). Analysis of myosin VI fragments has revealed that the globular tail region alone is sufficient to target myosin VI to uncoated vesicles (Aschenbrenner et al.,2003).

Cultured epithelial cells overexpressing the myosin VI globular tail region exhibit a dramatic reduction in transferrin uptake (∼20% of normal)(Aschenbrenner et al., 2003)confirming that the shorter myosin VI isoform also has a role in endocytosis. Remarkably, under these conditions initial rates of transferrin uptake into clathrin-coated vesicles are normal and instead the block is at the uncoated vesicle stage of endocytosis (Aschenbrenner et al., 2003). The uncoated vesicles remained 'stuck' at the cell periphery, dramatically delaying delivery of the transferrin cargo to the early endosome. Therefore, depending on the splice version present, myosin VI may act at an early or late stage of endocytosis.

The tail domain of myosin VI must associate with distinct cargo proteins that differentially localize the motor to clathrin-coated pits and/or uncoated endocytic vesicles. Three distinct linker proteins have been identified that associate with the tail of myosin VI: disabled 2 (Dab2), GAIP-interacting protein-C-terminus (GIPC) and synapse-associated protein 97 (SAP97).

Dab2 (also called DOC-2) is a putative tumor suppressor protein implicated in cell surface receptor turnover, endocytosis and cell signaling pathways. It is a complex molecule containing several well-characterized protein-binding motifs (Fig. 4). Near its N-terminus is a phosphotyrosine-binding (PTB) domain, which binds to multiple cell-surface receptors of the low-density lipoprotein receptor (LDLR) family,all of which contain a conserved NPXY motif(Morris and Cooper, 2001; Oleinikov et al., 2000). PTB domains are structurally similar to plekstrin homology (PH) domains and, like PH domains, the PTB domain of Dab2 binds to phosphoinositides and Dab2 can simultaneously associate with NPXY-containing proteins and phosphoinositide-containing lipids (Mishra et al., 2002).

Centrally located in Dab2 are a series of DPF motifs(Fig. 4), binding sites for the clathrin adapter AP-2, and these motifs are sufficient for targeting of Dab2 to clathrin-coated pits (Morris and Cooper, 2001). Also in this region are type I and type II binding sites for clathrin heavy chain (Mishra et al., 2002). The presence of these binding sites implicates Dab2 in regulating clathrin-coated vesicle formation. Indeed, an N-terminal fragment containing the PTB and the central AP-2/clathrin-binding domain is sufficient to initiate the formation of clathrin-coated vesicles from phosphoinositide-containing lipids in vitro, a process that is further accelerated in the presence of AP-2 adapters(Mishra et al., 2002). Dab2 also contains five NPF motifs, which in other proteins are sufficient for association with Eps15 homology (EH) domains found in a variety of accessory proteins involved in endocytosis. Therefore, Dab2 exhibits all the features characteristic of an endocytic adapter protein. Because it can also associate with LDLR family members, Dab2 may be involved in linking specific cargo with clathrin polymerization on the membrane.

Analysis of Dab2-knockout mice has confirmed that Dab2 functions in endocytosis. Renal proximal tubule cells from these knockout mice have fewer clathrin-coated pits and show defects in amino acid and vitamin uptake, a characteristic of defects in megalin endocytosis(Morris et al., 2002b). Megalin is a member of the LDLR gene family and associates directly with the Dab2 PTB domain (Oleinikov et al.,2000).

Dab2 associates with myosin VI through its C-terminal serine- and proline-rich region (Fig. 4)and binds to the C-terminal globular tail of myosin VI in vitro(Inoue et al., 2002; Morris et al., 2002a). Overexpression of Dab2 reorganizes surface AP-2, and myosin VI that contains the splice insert is recruited to these structures(Morris et al., 2002a). Therefore, Dab2 probably serves as the bridge that links the longer form of myosin VI to clathrin-coated pits. Indeed, antibodies to myosin VI co-immunoprecipitate Dab2, AP-2 and megalin from the proximal tubule,confirming the in vivo association of these proteins (Biemesderfer et al.,abstract). It is interesting to speculate that Dab2 might recruit myosin VI to ligand-bound megalin; this complex would then be primed to transport the ligand-bound receptor down the microvillus, where it would be anchored into clathrin-coated pits through association of Dab2 with clathrin, AP-2, and other accessory proteins.

One unanswered question is how the association of Dab2 with myosin VI is regulated. In in vitro binding assays, Dab2 can associate with both splice forms of myosin VI (Inoue et al.,2002; Morris et al.,2002a); however in vivo, the tail insert is required for targeting to clathrin-coated pits (Buss et al.,2001; Morris et al.,2002a). Constructs lacking this insert do not target to clathrin-coated pits even if Dab2 is present(Aschenbrenner et al., 2003). An as-yet-uncharacterized mechanism must exist that regulates myosin VI targeting to Dab2 in clathrin-coated pits, specifically recruiting the longer splice version. Because Dab2 is a linker protein capable of multiple simultaneous associations, perhaps another protein in the complex binds the insert sequence and confers this specificity. Alternatively, splicing or differential phosphorylation of Dab2 might regulate the integration of myosin VI into Dab2-containing clathrin-coated pits. Dab2 was first characterized as an alternatively spliced mitogen-regulated phosphoprotein with two predominant splice forms (Xu et al., 1995)(Fig. 4). Both identified splice forms of Dab2 can associate with myosin VI(Inoue et al., 2002; Morris et al., 2002a). However, the shorter Dab2 isoform might associate with AP-2 and clathin to a lesser extent because the region containing the important association motifs is missing (Mishra et al.,2002; Morris et al.,2002a).

GIPC is a PDZ-domain-containing protein known under a variety of monickers depending on the yeast-two hybrid screen bait used to identify it(Fig. 4). Its centrally located PDZ domain binds to proteins that have at their C-termini either a conserved type I PDZ-binding site (S/T)-X-(V/A)(Songyang et al., 1997), or a similar C-terminal sequence such as S-Y-S. This binding flexibility may explain the abundance of published GIPC associations.

The list of identified binding partners for GIPC includes many transmembrane proteins, such as multiple members of the LDL receptor family(e.g. megalin) (Gotthardt et al.,2000), the glucose transporter GLUT1C(Bunn et al., 1999), receptor tyrosine kinases [insulin-like growth factor-1 (IGF-1) receptor(Ligensa et al., 2001); TrkA and TrkB (Lou et al., 2001)],and the receptor serine/threonine kinase transforming growth factor β(TGFβ) receptor type III (Blobe et al., 2001). It has also been identified as a binding partner for many different cell surface molecules involved in adhesion, including 5T4 [a protein highly expressed in transformed cells that correlates with metastatic phenotypes (Awan et al.,2002)], integrins α5, α6a and α6b(El Mourabit et al., 2002; Tani and Mercurio, 2001), the adhesion regulator syndecan-4 (Gao et al.,2000), the semaphorins M-SemaF and SemC(Wang et al., 1999) and the semaphorin receptor neuropilin-1 (Cai and Reed, 1999). Although this list continues to grow, there is little in vivo evidence for association of GIPC with any of these proteins at the plasma membrane.

There is evidence for a role for GIPC in trafficking of transmembrane proteins through the Golgi stacks, however. GIPC associates with GAIP, a membrane-anchored GTPase-activating protein for G_αi3 subunits(De Vries et al., 1998b). GAIP localizes to clathrin-coated vesicles in the Golgi region, which implicates GIPC in membrane trafficking (De Vries et al., 1998a). GIPC also associates with gp75 tyrosinase related protein 1, a melanosomal membrane protein(Liu et al., 2001), but only with newly synthesized gp75 as it traverses the Golgi. Perhaps GIPC functions in the sorting of the other transmembrane receptors. It may also be involved in recruiting myosin VI to the Golgi, given that a fraction of myosin VI is reported to be Golgi associated (Buss et al., 1998).

Myosin VI was first identified as a binding partner of GIPC in a yeast two-hybrid screen (Bunn et al.,1999). The precise myosin VI-binding region on GIPC was not defined; however the PDZ domain of GIPC was found to not be sufficient for binding in vitro. GIPC is present on small vesicles near the plasma membrane in cultured cell lines (De Vries et al.,1998b) and myosin VI colocalizes with GIPC on these peripherally located vesicles (Aschenbrenner et al.,2003). Pulse-chase experiments confirmed that the vesicles are uncoated transferrin-containing endocytic vesicles, implicating GIPC in endocytosis (Aschenbrenner et al.,2003). In vivo, GIPC is enriched at both the clathrin-rich invaginations and the endocytic compartments found between microvilli in proximal tubule kidney cells, where it overlaps with GAIP(Lou et al., 2002), clathrin,AP-2 and myosin VI (Biemesderfer et al.,2002). Therefore, GIPC, in common with Dab2, might associate with myosin VI to cluster megalin or other receptors that have a type I PDZ-binding motif into clathrin-coated intermicrovillar regions. Unlike Dab2, however,GIPC can remain associated with myosin VI after the clathrin-coated vesicle is formed and uncoated, perhaps serving a role in later stages of vesicle trafficking.

SAP97 is a member of the PSD-95 family of membrane-associated guanylate kinase homologues (MAGUKs) (reviewed by Fujita and Kurachi, 2000). Unlike other SAPs, SAP97 is also expressed in non-neuronal cells and is present at cadherin-based cell-cell adhesions in epithelial cells(Muller et al., 1995; Reuver and Garner, 1998). It has three centrally located PDZ domains, as well as a Src-Homology 3 (SH3)domain and a C-terminal guanylate kinase (GUK) homology domain(Fig. 4), all domains classically involved in protein-protein interactions. The N-terminal domain of SAP97 is required for targeting of SAP97 to adhesion sites in epithelial cells(Wu et al., 1998). This domain associates with multiple binding partners, including three MAGUK scaffolding proteins [Lin-2, DLG2 and DLG3 (Karnak et al., 2002)] and myosin VI (Wu et al., 2002). The association with the MAGUKs likely mediates the targeting of SAP97 to adhesion sites in epithelial cells. An association between myosin VI and SAP97 is not seen in epithelial cells(Karnak et al., 2002; Wu et al., 2002), and evidence for a SAP97 - myosin VI association has only been reported in brain(Wu et al., 2002).

SAP97 is present throughout neurons as well as at the synapse. It is implicated in localization of the AMPA-type glutamate receptor subunit, GluR1. The first PDZ domain of SAP97 associates with a type I PDZ-binding motif found at the C-terminus of GluR1 (Leonard et al., 1998) and facilitates its trafficking through the Golgi to the plasma membrane (Sans et al.,2001). Since a fraction of cellular myosin VI resides in the Golgi(Buss et al., 1998), this could be associated with SAP97. Although SAP97 is a synaptic protein, thus far there is no evidence for association of SAP97 with GluR1 or myosin VI at the synapse. Moreover, there is also as yet no evidence for a role for SAP97 in endocytosis.

Dab2, GIPC or both may mediate recruitment of myosin VI to clathrin-coated pits, but what function is myosin VI serving here? The answer probably lies with actin, which has several potential roles in early stages of endocytosis(Apodaca, 2001; Qualmann and Kessels, 2002; Qualmann et al., 2000)(Fig. 2B), each of which could require myosin.

By associating with both actin and linker proteins, myosin VI might cluster receptors onto actin networks and thereby spatially organize the endocytic machinery. Such an arrangement would be particularly important in the intermicrovillar clathrin-rich endocytic regions of polarized epithelial cells, regions rich in myosin VI. In cultured cells, clathrin-coated pits are often aligned with the underlying actin cytoskeleton, particularly on basal cell surfaces (Puszkin et al.,1982). Movement of clathrin-coated pits within the plasma membrane has also been shown to require actin(Gaidarov et al., 1999). These mechanisms for clathrin-coated pit positioning within the plasma membrane have not been characterized but could involve a myosin such as myosin VI.

Alternatively, rather than having a strictly structural function, as a two-headed motor myosin VI may provide the force necessary for deformation of the plasma membrane seen at sites of pits or serve as a force generator during or after vesicle fission. In support of the latter hypothesis, overexpression of myosin VI tail fragments containing the tail insert does not alter the morphology of the clathrin-coated pits but does cause a defect in clathrin-coated vesicle formation (Buss et al., 2001). Higher-resolution analysis of clathrin-coated vesicle formation should distinguish between these two potential functions.

Studies of other actin-binding proteins implicated in endocytosis suggest that a mechanism to dissolve the cortical actin barrier is essential for endocytosis in some cell types, and this may involve myosin VI (reviewed by Qualmann and Kessels, 2002). Specifically, the actin-spectrin network must be dissolved for coated-pit budding in fibroblasts. The loss of spectrin depends on cleavage by an activated calpain protease, a calmodulin-dependent enzyme(Kamal et al., 1998). As a calmodulin-associated protein, myosin VI might be involved.

Evidence also indicates a need for actin polymerization during endocytosis(Qualmann and Kessels, 2002; Qualmann et al., 2000). Several actin-binding proteins that can recruit the polymerization machinery have been implicated in early steps of endocytosis(Olazabal and Machesky, 2001; Qualmann and Kessels, 2002),and actin polymerization has been observed at the neck of invaginating clathrin-coated pits (Merrifield et al.,2002). Actin polymerization might accelerate vesicle fission or in the creation of an actin 'tail' that provides the first push moving the vesicle away from the plasma membrane surface. Studies that focus on the budding step have shown that drugs that depolymerize actin have little effect on endocytosis in permeabilized fibroblasts or nonpolarized cells in vitro(Fujimoto et al., 2000; Lamaze et al., 1997), but significantly affect clathrin-mediated uptake in hepatoma cells and enterocytes (Durrbach et al.,1996; Jackman et al.,1994). Therefore, there is a potential role for polymerization in apical endocytosis particularly in polarized cells.

How does myosin VI fit into existing models for actin polymerization in clathrin-coated vesicle formation? Studies of the Drosophila myosin VI homologue, Jaguar (also known as 95F myosin), have implicated it in an actin polymerization process that occurs during spermatogenesis(Rogat and Miller, 2002). Jaguar is highly homologous to mammalian myosin VI, with the highest levels of sequence similarity being in the cargo binding tail domains suggesting conserved function (Hasson and Mooseker,1994). Jaguar mutations affect testes myosin VI gene expression and jaguar flies have a defect in the individualization stage of spermatogenesis. During this stage, membranes are laid down between each spermatid, separating each from its neighbors. A cone of actin precedes the addition of membrane, and myosin VI is enriched at the leading edge of this cone. Myosin VI is required at this position to bring in the actin polymerization machinery, including the ARP2/3 complex(Hicks et al., 1999; Rogat and Miller, 2002). Given the high levels of homology between fly and mammalian myosin VI, mammalian myosin VI might similarly be required to recruit the actin polymerization machinery to the newly forming clathrin-coated vesicle during endocytosis.

Because overexpression of the tail domain of myosin VI (lacking any spliced inserts) causes an accumulation of uncoated vesicles containing endocytosed transferrin, myosin VI might be involved in the movement of newly formed vesicles inwards to the early endosome(Fig. 2C)(Aschenbrenner et al., 2003). Under these conditions, the vesicles are competent for fusion with the early endosome and contain the fusion factor Rab5, but remain trapped in the actin meshwork at the periphery of the cell(Aschenbrenner et al., 2003). What is the myosin's role? It might simply transport the vesicles along actin filaments. This type of model has been suggested for other myosins involved in membrane trafficking (reviewed by Langford, 2002; Tuxworth and Titus, 2000),although, in all these cases, the myosin is transporting its cargo towards the cell periphery. Although straightforward, a simple transport model for myosin VI may not be sufficient to explain the dramatic delay in vesicle trafficking seen upon disruption of myosin VI function. The actin filaments within the cortical actin network found at the cell periphery are not all oriented with pointed ends inwards, as is seen at the plasma membrane. Therefore direct transport inward by myosin VI may not be feasible, and other possible roles for the motor must be considered.

Myosin VI might facilitate actin reorganization necessary for fusion of vesicles with the early endosome. Kinetic studies of myosin VI suggest that during most of its duty cycle it maintains a tight association with F-actin(De La Cruz et al., 2001). Because of this property, as a two-headed myosin, myosin VI could potentially associate with two distinct filaments and mediate filament sliding necessary to extricate the vesicle from the dense actin mesh found in peripheral regions of cells and the terminal web. Such rearrangements may be necessary for the fusion machinery to access the early endosome.

Alternatively, as described in the previous section, myosin VI could act as a regulator of actin dynamics. Actin polymerization has been implicated as a mechanism for movement of endocytic vesicles and actin comet tails have been seen on endocytic vesicles in mast cells(Merrifield et al., 1999). Therefore, myosin VI may function in the movement of endocytic vesicles by recruiting components necessary for actin polymerization, thereby pushing the uncoated vesicles out of the peripheral actin-rich domains and towards the early endosome.

Ultimately, a defect in any of these three processes could explain the accumulation of vesicles seen at the cell periphery after disruption of myosin VI function and further studies will be required to distinguish between them.

Analysis of phenotypes associated with myosin VI mutations has not directly implicated myosin VI in endocytosis, but does suggest roles in actin cytoskeleton assembly, and membrane trafficking. In mammals, myosin VI mutations cause profound neurosensory deafness and balance disorders(Avraham et al., 1995; Melchionda et al., 2001). Both phenotypes are due to defects in the inner ear hair cells, a polarized cell type with unique actin-based projections called stereocilia. Stereocilia are similar in structure to microvilli and are required for sensing sound and balance signals. Myosin VI is found at the base of the stereocilia in an actin rich structure called the cuticular plate, occupying a position similar to that seen in kidney microvilli (Hasson et al., 1997) (Fig. 1). Snell's waltzer (sv) mice, which lack myosin VI, have defects in stereocilia development (Self et al., 1999). Although microvilli are initially present on the surface of the hair cells, they do not go through the usual elongation and widening steps associated with stereocilia development and instead aberrant actin-containing protrusions are assembled on the hair cell surface. This specific developmental defect could be due to one of the many potential actin-associated activities of myosin VI(Fig. 2) including errors in actin polymerization or defects in the coordinated regulation of membrane trafficking.

In the inner ear, unlike the kidney, endocytosis does not occur between the actin projections of the stereocilia. Instead, endocytosis occurs in an apical domain outside the actin-rich stereocilia and cuticular plate called the pericuticular necklace (Hasson et al.,1997; Kachar et al.,1997). In keeping with a role in endocytosis, myosin VI is enriched within this necklace; however studies of hair cell fluid phase uptake in sv mice suggest that unregulated endocytosis at least is normal in these animals (Self et al.,1999). Analysis of clathrin-mediated uptake has not been evaluated in sv mice, but this result is in keeping with the observation that fluid-phase uptake is not affected when dominant-negative versions of myosin VI are expressed in cultured cells(Aschenbrenner et al., 2003; Buss et al., 2001). Therefore,myosin VI may have regulated functions in hair-cell endocytosis, or it may have other functions distinct from those proposed here in membrane trafficking. For example, studies in C. elegans have implicated myosin VI in the asymmetric movement of organelles seen during spermatogenesis(Kelleher et al., 2000). Furthermore, studies in Drosophila have similarly implicated myosin VI in the directed movement of intracellular particles, presumed to be membrane vesicles, during the syncytial blastoderm stage of embryogenesis(Mermall et al., 1994) and during oogenesis (Bohrmann,1997). Therefore, myosin VI may participate in the movement of many types of membrane organelles along the actin cytoskeleton and not just the vesicle populations discussed here.

Studies in a variety of organisms and cell types have revealed universal roles for myosin VI in membrane trafficking events. Recent data using mammalian cultured cell models has suggested that myosin VI plays a role in clathrin-coated vesicle formation and the trafficking of uncoated nascent vesicles. It is likely that in both processes, myosin VI plays an accessory role, perhaps increasing the efficiency of endocytosis. Myosin VI cannot play an indispensible role in either process as sv mice are viable and fertile and do not exhibit defects in endocytic function at a gross level(Avraham et al., 1995). Therefore, mechanisms must be in place in mammals to compensate for the lack of myosin VI, perhaps recruiting another as-yet-unidentified pointed-end directed myosin.

The fact that myosin VI has been implicated in two very distinct steps of the endocytic process is remarkable: depending on the cell type myosin VI is either recruited to uncoated vesicles or clathrin-coated pits. The difference in myosin VI targeting may be due to local differences in the actin cytoskeleton, to alternative splicing or to differences in components that bind to the C-terminal tail domain of myosin VI. Another possibility is that the differences in myosin VI targeting reflect different signaling pathways that are activated to recruit myosin VI to distinct regions. The best such candidates are p21-activated kinase (PAK) family members, which have been shown to phosphorylate myosin VI in vitro(Buss et al., 1998; Yoshimura et al., 2001). Myosin VI is indeed phosphorylated in vivo(Buss et al., 1998) (our unpublished data) but the significance of myosin-VI phosphorylation remains to be studied.