Increasing evidence from a variety of cell types has highlighted the importance of the actin cytoskeleton during endocytosis. No longer is actin viewed as a passive barrier that must be removed to allow endocytosis to proceed. Rather, actin structures are dynamically organised to assist the remodelling of the cell surface to allow inward movement of vesicles. The majority of our mechanistic insight into the role of actin in endocytosis has come from studies in budding yeast. Although endocytosis in mammalian cells is clearly more complex and subject to a greater array of regulatory signals, recent advances have revealed actin, and actin-regulatory proteins, to be present at endocytic sites. Furthermore, live cell imaging indicates that spatiotemporal aspects of actin recruitment and vesicle formation are likely to be conserved across eukaryotic evolution.
Endocytosis is the process through which the plasma membrane invaginates into the cell, resulting in the production of a vesicle that is then able to fuse with endosomes and enter the endo-lysosomal membrane system. Endocytosis is required for recycling of plasma membrane lipids and trafficking proteins, and for uptake or downregulation of cell-surface receptors. The endocytic process can also be exploited by pathogens and toxins to allow their entry into cells. Uptake of material by endocytosis of clathrin-coated vesicles has been visualised by electron microscopy for over 40 years (Roth and Porter, 1964). Clathrin-coated pits assemble at the plasma membrane, select cargo, invaginate and pinch off. Clathrin-coated vesicles then move into the cell, lose their coat and the uncoated endocytic vesicles fuse with the early endosome to deliver their content. In addition to clathrin-coated endocytosis, some eukaryotic cells are able to internalise external material by an engulfment process called phagocytosis, and more recently non-clathrin-coated-vesicle internalisation processes have been identified.
Actin is an ubiquitous eukaryotic protein that is able to form dynamic polar filamentous structures (F-actin), which constitute a major part of the cell cytoskeleton. F-actin associates with a plethora of binding proteins that regulate the rate of filament assembly and disassembly, as well as organising the filaments into higher-order structures, such as bundles or crosslinked meshworks (Winder and Ayscough, 2005). Many studies have observed that actin localises to regions underlying the plasma membrane of cells. However, it is from studies of budding yeast that the clearest data supporting a role for actin in endocytosis have emerged. Indeed, until recently this role for actin was generally assumed to be a rather specialised function for yeast. Only since the advent of high-resolution, dual-labelling and real-time fluorescence microscopy has the clear participation of actin in mammalian cell endocytosis been observed. This has led the way for experiments that now indicate that the role of actin in endocytosis is likely to have been evolutionarily conserved.
In this Commentary, we limit our discussion to the role of actin in clathrin-mediated endocytosis. This is the major uptake pathway of many mammalian cells. However, the significance of other endocytic pathways is becoming increasingly recognised (Conner and Schmid, 2003). In all cases, actin has been implicated in the mechanics of uptake, although to varying degrees. In the case of phagocytosis, the importance of actin in pseudopod extension and engulfment has been known for many years. However, our knowledge of the mechanistic basis for actin involvement in macropinocytosis and non-clathrin-mediated uptake is currently limited.
Below, we outline the current understanding of the function of actin during endocytosis and highlight how actin is regulated to allow the formation and movement of endocytic vesicles from the plasma membrane. Initially we focus on the data from yeast, which form the basis for most of our current understanding of the role of actin during the process. We then consider endocytosis in mammalian cells, and discuss recent advances that demonstrate the importance of the actin cytoskeleton in this context.
Actin in yeast and the mechanism of endocytosis
Actively growing budding yeast contains three distinct actin structures that are visible by fluorescence microscopy of Rhodamine-phalloidin-stained cells: cortical actin patches, which are now thought to represent sites of endocytosis; actin cables, which are crucial for polarity and for movement of exocytic vesicles; and a contractile actin ring, which plays a role in cytokinesis (Adams and Pringle, 1984).
The first evidence linking actin to the endocytic process was published 13 years ago. A genetic screen for yeast mutants exhibiting defective uptake of a fluid-phase marker revealed that, as well as displaying defective endocytosis, these mutants also contain aberrant actin (Kübler and Riezman, 1993; Raths et al., 1993). Over subsequent years, links between proteins of cortical actin patches and the ability to effectively endocytose became increasingly apparent. However, although >30 proteins were identified and characterised as components of the cortical actin patches, how the actin cytoskeleton is involved in the process remained unclear (reviewed in Engqvist-Goldstein and Drubin, 2003).
Key advances have come from studies of the behaviour of actin patch proteins in real-time. Several studies have assessed the behaviour and lifetime of these patches (Doyle and Botstein, 1996; Waddle et al., 1996; Warren et al., 2002). Such studies show that a patch is formed at the plasma membrane and after a few seconds it moves rapidly into the cell (Fig. 1). However, other proteins that regulate actin appear to have distinct behaviours, often remaining relatively non-motile at the membrane or showing just small inward movement (Warren et al., 2002). How this behaviour relates to the process of endocytosis was explained in a very elegant study in 2003, when Kaksonen and colleagues used dual-colour, real-time fluorescence microscopy to study kinetic aspects of the endocytic process (Kaksonen et al., 2003). This study led them to propose a model linking the recruitment of actin to the endocytic process. Essentially, endocytic vesicle coat proteins are thought to be recruited to sites of the plasma membrane, where they act to deform the membrane, causing invagination. Actin is then recruited to assist the later stages of vesicle formation and the rapid movement of the vesicle away from the membrane.
The laboratories of David Drubin and others have subsequently added to this core data to propose a clear mechanism through which organisation and regulation of actin is integral to the endocytic process (Jonsdottir and Li, 2004; Kaksonen et al., 2003; Kaksonen et al., 2005; Newpher et al., 2005; Soulard et al., 2005; Warren et al., 2002). As depicted in Fig. 2, the process involves distinct stages of non-motile, slow-moving and fast-moving patches that can be defined by the presence of discrete protein modules. The four main modules that assemble at endocytic sites in a coordinated and sequential manner are: the endocytic coat complex, the actin network growth machinery, the actin dynamic regulation module and the vesicle scission module.
Actin regulation during endocytosis in budding yeast
The endocytic coat and actin nucleation
Prior to actin recruitment, the site of vesicle formation is determined and the initial stages of coat formation are completed. The selection and establishment of an endocytic site is achieved by a group of proteins referred to as the `endocytic coat complex'. This group contains many proteins, which perform a range of functions. Clathrin is able to form a basket-like network around a forming vesicle that is thought to assist in maintaining curvature. Other proteins such as Ent1p and End3p contain ENTH domains that interact with specific lipids, such as phosphoinositol (4,5)-bisphosphate [(PtdIns(4,5)P2], and are proposed to drive initial stages of membrane curvature. A third group of proteins including Sla1p, Sla2p and Pan1p contain multiple interaction domains. Sla1p binds endocytic cargo and other coat proteins and negatively regulates the Arp2/3 activator Las17p; Sla2p contains a membrane-interaction domain, a central domain that binds other coat components and a C-terminal actin-binding domain; Pan1p binds coat proteins End3p and Yap1801/2, as well as functioning as an activator of Arp2/3 (see Table 1 for known mammalian orthologues of these proteins). These proteins associate early and are visualised as a non-motile patch that later moves a small distance into the cell. This movement is proposed to occur during the invagination and scission steps, after which these proteins disassemble from the vesicle.
What then is the role of actin in the process, and how is it regulated? The importance of F-actin in endocytosis is clearly illustrated by the effect of addition of the actin-monomer-sequestering drug latrunculin A. Within 5 minutes of its addition, actin patches are no longer visible and endocytosis is completely abrogated (Ayscough et al., 1997; Morton et al., 2000). The stage in the endocytic process at which actin is observed is concomitant with onset of slow vesicle movement. This is the point at which the yeast WASP orthologue Las17p and the type I myosins (Myo3p and Myo5p) begin to nucleate actin filaments through activation of the Arp2/3 complex (Sun et al., 2006). This group of proteins has been called the `actin network growth machinery'.
Las17p is present on the membrane for a prolonged period, whereas type I myosins and several regulatory proteins (Vrp1p, Bzz1p and Bbc1p) arrive just prior to actin (see Table 1 for the mammalian orthologues). Live cell imaging reveals that these proteins remain in a non-motile complex at the plasma membrane throughout the invagination and scission process. The current idea is that Las17p (WASP) interacts with coat components including Sla1p, which maintain it in an inactive state. Las17p subsequently recruits Vrp1p (the WIP orthologue) and Bzz1p (the syndapin orthologue) (Sirotkin et al., 2005; Soulard et al., 2002). Bzz1p relieves the inhibition of Las17p by Sla1p to allow initiation of actin nucleation at the endocytic site. Vrp1p and F-actin recruit Myo3p and Myo5p to the site and the actin-nucleation-promotion activity of Myo5p is then activated by Vrp1p (Sun et al., 2006; Jonsdottir and Li, 2004; Soulard et al., 2005; Soulard et al., 2002). Fine-tuning of the level of actin nucleation required is proposed to be the function of Bbc1p, Abp1p and coronin (Humphries et al., 2002; Rodal et al., 2003; Rodal et al., 2005). As illustrated in Fig. 2, filament formation then drives membrane invagination and `pushes' the forming vesicle away from the plasma membrane. After scission, these proteins are observed to disassemble from the endocytic site.
The actin dynamic regulation machinery
Soon after nucleation, actin filaments are capped and bundled. Real-time imaging reveals that the proteins responsible for this activity show similar behaviour to actin itself. They arrive at the endocytic site at the slow movement (vesicle invagination) step; then they move quickly into the cell, remaining with the vesicle after coat protein disassembly. The only characterised capping protein in S. cerevisiae is a heterodimer encoded by the CAP1 and CAP2 genes. It is highly homologous to the Capα and Capβ proteins that form the CapZ heterodimer in mammalian cells. Capping protein is proposed to limit addition of monomers to newly synthesized filaments, facilitating rapid filament growth close to the cell cortex (Kim et al., 2006; Kim et al., 2004). Δcap1Δcap2 cells show a reduced rate of patch movement (Kaksonen et al., 2005). Interestingly, the effect on endocytosis is not highly penetrant, which indicates that either capping is a relatively minor participant in the process or that other capping proteins exist in yeast.
Following actin polymerisation, the filaments are bound together by the activity of Sac6p/fimbrin a highly conserved actin-bundling protein. In the absence of Sac6p, actin polymerises at endocytic sites, but there is no internalisation (Kaksonen et al., 2005; Kübler and Riezman, 1993). Actin bundles are stronger than single filaments and are thought to provide the strength required to allow generation of sufficient force to allow inward movement of the forming vesicle. Other proteins, including Scp1p, associate with actin at this stage and are likely to provide further stability to the actin network (Winder et al., 2003). However, the functions of many of these proteins await further analysis of their effects on actin dynamics and the endocytic process.
The filaments must then be depolymerised to release monomers to allow further rounds of endocytosis. The best characterised depolymerisation factor in yeast is cofilin, which, in common with many of its orthologues in other eukaryotes, is essential (Iida et al., 1993; Lappalainen and Drubin, 1997; Moon et al., 1993). Mutant alleles of cofilin that cause actin patches to grow larger have been generated and there is an associated defect in endocytosis (Lappalainen and Drubin, 1997). The importance of a dynamic actin cytoskeleton rather than the presence of F-actin alone is also illustrated by experiments involving addition of the drug jasplakinolide, which stabilises F-actin in cells; this completely abrogates endocytosis in yeast (Ayscough, 2000)
Mutations in at least 20 proteins are known to affect the function of actin during this process, and the presence of a recognised level of redundancy among the proteins involved means that this number is likely to be higher. In addition to proteins that directly bind to and affect actin dynamics, there is a further level of control exerted over the actin-binding proteins themselves. One of the best characterised of these mechanisms is phosphoregulation. The key kinases that have been studied are Ark1p and Prk1p, which are related to the mammalian kinases AAK1 and GAK (Smythe and Ayscough, 2003). In yeast, these phosphorylate several components of the endocytic coat, including Pan1p, Ent1p and Sla1p (Ayscough, 2005; Huang et al., 2003; Watson et al., 2001; Zeng et al., 2001). Phosphorylation of Pan1p has been shown to inhibit its ability to activate Arp2/3-dependent actin polymerisation (Duncan et al., 2001; Toshima et al., 2005), whereas phosphorylation of Sla1p is required for its release from the coat (Zeng et al., 2001). If Sla1p is not removed from the coat, actin turnover becomes reduced and endocytosis is inhibited. The kinases are recruited to the sites of endocytosis by another actin-binding protein, Abp1p (Fazi et al., 2002).
Actin and vesicle scission
Invagination is followed by vesicle scission, and then movement of the vesicle away from the membrane. The role of actin in these stages is less well characterised. The vesicle scission module contains the two yeast amphiphysin proteins Rvs161p and Rvs167p. These proteins contain BAR (Bin-Amphiphysin-Rvs) domains that bind to and tubulate membranes in vitro (Peter et al., 2004). Tubulation is then suggested to facilitate the scission process. The yeast amphiphysins have been shown to bind to a number of known endocytic proteins, including Las17p, Sla1p and actin itself (Lombardi and Riezman, 2001; Friesen et al., 2003; Madania et al., 1999; Stamenova et al., 2004). The proteins arrive after actin, remain immobile for a few seconds, then show transient movement into the cell. In rvs161 or rvs167 deletion mutants, cells show frequent retraction of the forming vesicle after an initial internalisation movement, which may suggest failed endocytosis.
It is not currently clear whether actin plays a direct role in scission. However, actin and associated proteins recruit amphiphysins and thus mark the site at which membrane tubulation should occur. It should also be noted that endocytosis can still occur in the absence of the amphiphysins, which suggests that other mechanisms for scission exist. One possibility is that other proteins, such as dynamin, are involved. Dynamin is a GTPase that plays a key role in the endocytic scission process in mammalian cells (Song and Schmid, 2003). Although dynamin relatives are present in yeast, there is no experimental support for a role for this family of proteins in endocytosis (Engqvist-Goldstein and Drubin, 2003). Alternatively, in the absence of amphiphysins, continued inward movement of the endocytic pit mediated by interactions between the coat and actin bundles could lead to neck elongation and eventually to pinching off. Very elongated vesicle necks are seen in a number of strains that possess mutations in genes encoding actin regulatory proteins (Jonsdottir and Li, 2004; Wendland et al., 1996), indicating the possible importance of actin at this scission step.
Actin and vesicle movement
Following scission, the vesicle is uncoated and moves away from the membrane until it fuses with an endosome. There are two possible roles for actin at this final stage: vesicles could move along actin cables; alternatively actin could be nucleated at the vesicle surface to facilitate their movement within the cell. Evidence for the former idea includes the demonstration that endocytic vesicles containing the lipophilic dye FM4-64 undergo linear movement along actin cables (Huckaba et al., 2004). Furthermore, a recent study of receptor-mediated uptake and trafficking of fluorescently labelled alpha-factor derivatives revealed that endocytic vesicles and endosomes move towards each other along the actin cables (Toshima et al., 2006). Genetic studies revealing links between proteins involved in formation of actin cables (e.g. formins) and endocytic proteins such as Sla2p lend further support to the idea that actin cables act as tracks to facilitate the endocytic process (Yoshiuchi et al., 2006). Clearly, we must now identify the proteins that link vesicles to these actin cables.
Currently, there is no clear evidence that actin nucleation drives vesicle movement. Such a process would be analagous to that used by Listeria where actin is polymerised at the bacterial surface and can be visualised as `comet-tails' within cells as the Listeria move. The presence of Abp1p and Pan1p on the newly formed vesicles, which are both activators of Arp2/3-dependent actin polymerisation (Duncan et al., 2001; Goode et al., 2001) might, however, mean that although the majority of the F-actin polymerisation drives the initial invagination process, a lower level of filament formation facilitates movement of vesicles away from the membrane.
Actin and endocytosis in mammalian cells
Until recently, the role(s) of actin in clathrin-mediated endocytosis in mammalian cells was difficult to define, although there were intriguing observations connecting actin and its binding partners to components of the endocytic machinery (Qualmann and Kessels, 2002). For example, in neuronal cells, actin filaments were visualised in close proximity to sites of endocytosis at the synapse (Shupliakov et al., 2002). However, treatment of cells with drugs that interfere with the actin cytoskeleton resulted in variable effects on endocytosis, which suggested at the very least that actin does not have a mandatory role in clathrin-mediated endocytosis and simply facilitates the process, depending on the context (Fujimoto et al., 2000). Recent studies, largely based on live-cell imaging, now provide more compelling evidence for a role for actin in endocytosis and progress has been made to elucidate the mechanistic basis of its involvement.
Association of actin with mammalian clathrin-coated vesicles
A major step forward definitively linking actin polymerisation to clathrin-coated vesicle formation came from the live-cell imaging studies of Merrifield and colleagues (Merrifield et al., 2002; Merrifield et al., 2005). These studies combined epifluorescence with total internal reflection microscopy (TIRF) to follow the internalisation of individual coated pits in living cells expressing a fluorescently tagged form of clathrin light chain (DsRed clathrin). The combination of these two approaches allowed the authors to distinguish the plasma-membrane-associated pool of clathrin from the bulk of the clathrin within the cell. They showed that transient recruitment of actin coincides with the inward movement of vesicles (Merrifield et al., 2002). Interestingly, actin is recruited during the internalisation of the majority of coated pits and vesicles, which indicates that its participation in mammalian endocytosis is widespread. Merrifield et al. also observed recruitment of the Arp2/3 complex to clathrin-coated pits, using this technique (Merrifield et al., 2004).
Further advances in imaging have allowed investigators to temporally distinguish components required for invagination from those required for scission. pH-sensitive probes have been developed that allow internalisation of cargo into individual clathrin-coated vesicles to be visualised, allowing us to follow the time course of invagination and identify the point of scission (Merrifield et al., 2005).
Actin regulation in the clathrin-coated-vesicle cycle
Treatment of cells with latrunculin A has shown that actin acts at multiple steps in the mammalian endocytic pathway. In living cells, the dynamic movement of clathrin spots and patches is inhibited in the presence of the drug (Merrifield et al., 2002; Yarar et al., 2005), and stage-specific assays in permeabilised cells indicate that invagination and scission are both affected by its sequestration of monomeric actin (Yarar et al., 2005).
The recruitment of actin during invagination is thought to provide the force that drives the invagination of the coated pit (Engqvist-Goldstein and Drubin, 2003; Merrifield, 2004; Yarar et al., 2005). In contrast to the situation in yeast, the large GTPase dynamin appears to play a key role integrating actin dynamics with endocytosis, interacting with numerous endocytic and actin-binding proteins (Schafer, 2004). Whereas dynamin has a well-established role in the scission of clathrin-coated vesicles (Song and Schmid, 2003), live cell imaging experiments have indicated that dynamin is recruited during invagination (Ehrlich et al., 2004; Merrifield et al., 2002), which was predicted on the basis of biochemical experiments (Gad et al., 2000; Hill et al., 2001). This places dynamin in the appropriate spatial and temporal position during the coated-vesicle cycle to recruit actin-binding proteins to the invaginating pit.
As in yeast, activators of the Arp2/3 complex are recruited to coated pits in mammalian cells through interactions with coat components. A likely candidate to initiate actin polymerisation at this early stage is N-WASP. This has been shown to be recruited to coated pits in living cells (Merrifield et al., 2002) and knocking it down by RNAi reduces the rate of epidermal growth factor receptor (EGFR) uptake (Benesch et al., 2005; Innocenti et al., 2005). N-WASP could potentially be recruited by a variety of SH3-domain-containing proteins – cortactin, endophilin, syndapin and intersectin – that enhance its ability to activate Arp2/3 and that also bind to dynamin. Syndapin and endophilin have also been implicated in recruitment of N-WASP during coated-vesicle formation (Kessels and Qualmann, 2002; Otsuki et al., 2003), whereas intersectin has been implicated in the N-WASP-mediated uptake of the T-cell antigen receptor in T cells (McGavin et al., 2001). Another Arp2/3 activator, cortactin, is also recruited to coated pits, attaining maximal levels at scission (Merrifield et al., 2005). Although cortactin is recruited during coated-pit invagination, experimental evidence suggests that it is unlikely to act in the initial recruitment of actin in the clathrin-coated vesicle cycle, because it needs to bind F-actin in order to effectively bind and activate the Arp2/3 complex. Cortactin thus probably has a role in the dynamic turnover of actin filaments during clathrin-coated vesicle formation (Zhu et al., 2005).
Mammalian Abp1p (mAbp1), like its yeast orthologue Abp1, provides another important link between the endocytic pathway and the actin cytoskeleton. Early studies showed that mAbp1 binds to dynamin through its SH3 domain and that the two proteins interact in vivo. Overexpression of the SH3 domain of mAbp1 has a dominant-negative effect on endocytosis (Kessels et al., 2001). Furthermore, studies in mAbp1-knockout mice reveal defects in the immune response as a result of impaired T-cell endocytosis (Han et al., 2005). Mice lacking mAbp1 also suffer from behavioural abnormalities and exhibit reduced rates of endocytosis and synaptic vesicle recycling. Interestingly, analysis of these cells shows that the absence of mAbp1 affects events following scission of coated vesicles, specifically causing an increase in the number of endosomal intermediates and a significant reduction in the rate of synaptic vesicle repriming (Connert et al., 2006).
Another family recently implicated in the dynamic interplay between the endocytic pathway and the actin cytoskeleton is the PCH/F-BAR protein family (Itoh et al., 2005; Tsujita et al., 2006). The F-BAR domain of these proteins allows them to bind to phosphoinositides and to bend and tubulate membranes in liposome-based assays. Several family members can bind dynamin and co-localise with dynamin and actin at the cell surface. Overexpression of the F-BAR proteins causes tubules to form from the plasma membrane, an effect accentuated by treatment of cells with latrunculin A. Co-overexpression of wild-type but not mutant dynamin inhibits tubulation, leading to the suggestion that both dynamin and the rigidity of the actin cytoskeleton limit the ability of these proteins to tubulate membranes and therefore co-operate to regulate the degree of invagination of clathrin-coated pits or other early endocytic intermediates. Syndapin I is also a member of this family. Instead of acting downstream of dynamin in the endocytic process, it might therefore use its F-BAR domain to sense membrane curvature and generate invagination; its activity could subsequently be curtailed by dynamin binding (Anggono et al., 2006). This hypothesis is partially based on the fact that syndapin I only has one SH3 domain. However, recent studies have shown that syndapin I dimerises in vivo, which would allow it to bind both dynamin and N-WASP simultaneously (Kessels and Qualmann, 2006).
Orthologues of the yeast proteins that cap, bundle and stabilise actin all exist in mammalian cells (Table 1) but their relevance in endocytosis mostly remains to be explored. Coronin regulates phagosome formation in mammalian cells, apparently by recruitment of Arp2/3 to sites of actin remodelling (Yan et al., 2005), and in the filamentous fungi Dictyostelium (de Hostos, 1999). Similarly fimbrin has been observed at the phagocytic cup and in macropinosomes in Dictyostelium but, because of apparent functional redundancy, it has not yet been possible to establish its precise role (Pikzack et al., 2005). Cofilin is also required for endocytosis in mammalian cells (Nishimura et al., 2006) but further studies will be required to define where and how it acts.
Recent studies have identified a conserved motif that mediates the interaction of members of the CIN85/CD2AP protein family with capping protein and shown that CD2AP inhibits the activity of capping protein in vitro (Bruck et al., 2006). CIN85 and CD2AP are probably the mammalian orthologues of Sla1p, although their primary roles may be cell-type specific – they appear to be required for specific cellular events (Dikic, 2002). For example, CIN85 acts as an adaptor/scaffolding molecule to recruit signalling molecules following the activation of receptor tyrosine kinases in HEK293T cells (Haglund et al., 2002) and CD2AP has very specific roles in the regulation of T-cell activation (Shaw, 2005) and kidney glomerular function (Wolf and Stahl, 2003). Both interact with components of the endocytic machinery and actin cytoskeleton – CIN85 interacts with endophilin, synaptojanin and Huntingtin-interacting protein 1 (HIP1) and Hip1-related protein (HIP1R) (Kowanetz et al., 2004); CD2AP interacts with Rab4, AP2, cortactin and actin (Cormont et al., 2003) – and disruption of this leads to reduced downregulation of the T-cell receptor in vivo (Hutchings et al., 2003). It will be interesting to investigate whether targeted disruption of the conserved motif linking this protein family to capping protein affects endocytosis in vivo.
HIP1R, the mammalian orthologue of Sla2p, provides a further link between endocytic vesicles and the actin cytoskeleton. HIP1R binds to clathrin light chains and actin and thus could bridge endocytic vesicles to the actin cytoskeleton (Chen and Brodsky, 2005; Legendre-Guillemin et al., 2005). Indeed, cells treated with siRNA to knock down HIP1R, and fibroblasts from mice lacking the related HIP1 protein, have reduced rates of coated-vesicle formation (Engqvist-Goldstein et al., 2004). Interestingly, however, in the absence of HIP1R, actin, dynamin and cortactin accumulate at sites of endocytosis, which suggests that HIP1R also modulates actin turnover at endocytic sites (Engqvist-Goldstein et al., 2004).
Actin as a facilitator of endocytosis
Although drugs such as latrunculin A can inhibit endocytosis in mammalian cells, they tend to reduce the rate of internalisation rather than blocking it completely. Indeed, endocytosis in some cell types is apparently unaffected by these treatments (Fujimoto et al., 2000). This strongly suggests that, in contrast to yeast, actin has a facilitatory rather than an essential role in endocytosis in mammalian cells. Actin could, however, become essential, depending on the context in which the cell finds itself. Subcellular localisation of signalling molecules frequently determines their output (Miaczynska et al., 2004). Furthermore, recent studies have suggested that there are subpopulations of clathrin-coated pits that select cargo (e.g. signalling molecules) destined to be degraded rapidly by the cell for delivery to a motile population of endosomes (Lakadamyali et al., 2006). A number of endocytic molecules act as `alternative' cargo adaptors that, either independently or in concert with the classical AP2 adaptor protein complex, select specific cargo for incorporation into clathrin-coated pits (Traub, 2003). Both HIP1R and HIP1 share characteristics of alternative adaptors in that they bind clathrin and phosphoinositides through their ANTH domains (Brett et al., 2006) and HIP1 may act as a cargo receptor for the AMPA receptor (Metzler et al., 2003). Cargo adaptors could thus determine the rate at which particular cargo traverses the endocytic pathway by modulation of actin dynamics.
Flux through the endocytic pathway must therefore play a key role in allowing cells to migrate, proliferate or differentiate during development and actin may well be differentially recruited by coated pits. In this context the diversity of proteins that can apparently link endocytic vesicles to the actin cytoskeleton in multicellular organisms may allow cells to respond to different environments by altering the rate at which cargo moves through the endocytic pathway. Certain proteins, such as syndapins and CIN85/CD2AP, may have additional cargo-specific roles. Indeed recent studies have indicated that syndapin is the relevant binding partner for dynamin in synaptic vesicle recycling (Anggono et al., 2006).
If the requirement for actin in endocytosis in mammalian cells increases in certain cellular contexts, then those proteins that link endocytic vesicles to the actin cytoskeleton should be subject to tight regulation. For example, the interaction between syndapin I and dynamin is modulated by phosphorylation (Anggono et al., 2006). Somewhat surprisingly, studies have not yet uncovered a major role for the Ark/Prk orthologues in mammalian cells in the connections between actin and endocytosis. AAK1 is likely to enhance recruitment of cargo into coated pits (Conner and Schmid, 2002; Ricotta et al., 2002), and the DnaJ domain rather than the kinase domain of GAK has been implicated in uncoating (Umeda et al., 2000; Ungewickell et al., 1997). However, many of the key proteins involved in the link between endocytosis and the actin cytoskeleton are phosphoproteins and so phosphorylation is probably a major regulatory mechanism.
Conclusions and future perspectives
Although the last 2-3 years have seen dramatic advances in our understanding of the endocytic process and the role of actin, studies will now need to refine the experimental systems to address the contributions of the plethora of non-essential components in both yeast and mammalian cells. The ability to measure more subtle differences in endocytic rates and to define the temporal order of recruitment of endocytic components will allow us to examine apparently redundant protein functions. These studies will be especially powerful when combined with siRNA knockdown and rescue approaches.
Fundamental questions raised by the recent work focus on the differences between constitutive and cargo-mediated endocytosis and the role of lipids in the regulation of actin-binding proteins during the invagination and scission stages of the process. Thus, a key issue is whether individual cargo molecules can directly influence actin dynamics to facilitate their uptake. Scaffolding molecules such as Sla1p in yeast and CIN85 in mammalian cells are good candidates to integrate signals for individual cargoes, and CIN85 has already been implicated in the uptake of specific cargoes including IgE, EGFR and the h-Met receptor (Molfetta et al., 2005; Petrelli et al., 2002; Soubeyran et al., 2002). Interestingly, myosin VI, which has been implicated in endocytosis from the apical membrane of polarised cells (Buss et al., 2002), interacts with the Dab2 `alternative' adaptor, which mediates the specific uptake of the LDL receptor family (Dance et al., 2004; Morris et al., 2002). This suggests a further molecular link between cargo uptake and modulation of actin dynamics. In yeast the recent report that myosin phosphorylation is required for ligand-induced but not for constitutive endocytosis of the pheromone receptor Ste2p (Grosshans et al., 2006) indicates that further levels of regulation are likely to be identified, possibly for many specific endocytic events.
Finally, researching the interplay between lipids and proteins will be vital to our understanding of actin's role in endocytosis. The importance of protein-lipid interactions for endocytosis is clear, and without the appropriate lipid environment, membrane curvature, which is crucial for vesicle formation, cannot occur. However, membrane lipids also regulate the activity of several actin-binding proteins, including capping protein (Amatruda and Cooper, 1992). In addition, the Arf family of small GTPases integrate signals to coordinately regulate membrane traffic and actin dynamics through responses to PtdIns(4,5)P2 levels in the membrane. Mammalian Arf6 functions at the plasma membrane to regulate both clathrin- and non-clathrin-mediated uptake whereas Arf1 coordinates trafficking through the trans-Golgi network, using many of the same molecular players (e.g. N-WASP and cortactin) (D'Souza-Schorey and Chavrier, 2006). Yeast Arf3p (the orthologue of Arf6) has also been implicated in the endocytic process (Costa et al., 2005), as have the synaptojanins, which deplete PIP2 levels possibly to facilitate the final stages of invagination and scission (Cremona et al., 1999; Bottcher et al., 2006; Stefan et al., 2005).
The advances in our mechanistic knowledge of endocytosis have come about from the use of multi-disciplinary approaches using both yeast and mammalian cells. These studies have now given us the robust foundation from which we can progress to investigate both fundamental and more specific questions concerning this essential and ubiquitous eukaryotic process.
We thank Ewald Hettema and Karen Smillie for critical reading of this manuscript and Iwona Smaczynsla-de Rooij for the Abp1p images used. K.R.A. is supported by a Medical Research Council (MRC) Senior Research Fellowship (G117/394). E.S. is supported by grants from the MRC (program grant: G0300452) and BBSRC (BBS/B/04072).
- Accepted September 6, 2006.
- © The Company of Biologists Limited 2006