Endocytosis, the process whereby the plasma membrane invaginates to form vesicles, is essential for bringing many substances into the cell and for membrane turnover. The mechanism driving clathrin-mediated endocytosis (CME) involves > 50 different protein components assembling at a single location on the plasma membrane in a temporally ordered and hierarchal pathway. These proteins perform precisely choreographed steps that promote receptor recognition and clustering, membrane remodeling, and force-generating actin-filament assembly and turnover to drive membrane invagination and vesicle scission. Many critical aspects of the CME mechanism are conserved from yeast to mammals and were first elucidated in yeast, demonstrating that it is a powerful system for studying endocytosis. In this review, we describe our current mechanistic understanding of each step in the process of yeast CME, and the essential roles played by actin polymerization at these sites, while providing a historical perspective of how the landscape has changed since the preceding version of the YeastBook was published 17 years ago (1997). Finally, we discuss the key unresolved issues and where future studies might be headed.
ENDOCYTOSIS is a key mechanism that regulates plasma membrane composition and thus fundamentally supports cellular physiology and homeostasis. It is the process by which cells collect extracellular material and plasma membrane-associated surface proteins, such as receptors, channels, and signaling proteins, and package them into vesicles that are pinched off and enter the cytosol (Figure 1). Internalized vesicles then fuse with other internal compartments so that their contents can be recycled or degraded.
In most steps of intracellular traffic, including endocytosis, proteins (referred to as “cargo”) move between intracellular organelles and the extracellular milieu via formation of vesicles or tubules. This process begins when transmembrane cargo recruit cytosolic proteins to the membrane surface. These proteins then assemble in a temporally ordered pattern suggesting layers or coats (Figure 1). The inner “adaptor” layer of coat proteins binds to sorting signals in the cytosolic portions of the transmembrane cargo proteins to collect and concentrate the cargo and also helps to recruit and organize subsequent outer coat layers. Among the earliest proteins recruited to sites of vesicle formation are the namesake “coat” proteins (e.g., clathrin, COPI, COPII), which polymerize into spherical cages of different sizes to promote vesicle budding. A final set of components acts on the membrane itself to facilitate scission and release of the vesicle into the cytosol (Figure 1).
Clathrin-mediated endocytosis (CME) is the best characterized of all the endocytic pathways and requires a large number of cytosolic proteins that concentrate transmembrane cargo proteins and form clathrin-coated vesicles (McMahon and Boucrot 2011). In CME, adaptor proteins recognize sorting signals in the cargo proteins and directly link the cargo to clathrin. The basic subunit of clathrin is a triskelion, which is a trimer of clathrin heavy chains, each with an associated clathrin light chain. The controlled formation of a dense and branched actin-filament network provides forces critical for driving subsequent membrane tubulation and vesicle scission (Figure 1). Although there is an absolute requirement for dynamic rearrangements of the actin cytoskeleton to support CME in yeast, there is a more variable requirement in animal cells, discussed below.
Our goal in this review is to summarize the current understanding of endocytosis and its relationship with actin dynamics in budding yeast, while highlighting the major advances that have been made since publication of chapters on this topic in the last edition of the YeastBook (Botstein et al. 1997; Kaiser et al. 1997), and finally to point out some of the remaining open questions in this field. We hope that this information will be useful to newcomers and experts alike and to those not directly working on endocytosis or actin biology. Looking back, it becomes strikingly evident that in 1997 our knowledge was in its infancy and that major leaps have since been made in our understanding of actin and endocytosis. There is now a wealth of information on the >50 components of endocytic sites, the temporal order of their arrival, their biochemical activities, physical interactions, and relative abundances, along with recent ultrastructural analysis clarifying the successive steps in endocytic internalization of vesicles from the plasma membrane. While much of this work was initially pioneered in Saccharomyces cerevisiae, many critical insights as well as important quantitative information (e.g., numbers of molecules per actin patch and mathematical modeling of actin networks; see below) have been contributed by studies in Schizosaccharomyces pombe (e.g., Berro et al. 2010; Sirotkin et al. 2010; Arasada and Pollard 2011). We attempt to compile all of this information together into a working model for the formation and internalization of endocytic vesicles. We apologize to all those whose work we did not highlight. Even in a review of this length, it was not possible to discuss this subject and its literature base comprehensively, and we were faced with difficult choices regarding where to focus. However, we refer readers to a series of excellent recent reviews on actin and endocytosis that provide complementary information and more in-depth discussion of specific areas that we do not (Boettner et al. 2012; Mooren et al. 2012; Weinberg and Drubin 2012; Idrissi and Geli 2014; Schmid et al. 2014).
Cortical actin patches represent the final stages of endocytic internalization
It is now clear that the plasma membrane-associated structures referred to as “cortical actin patches” correspond to sites of endocytosis in the late (actin-dependent) stages of vesicle formation. However, in 1997, the equivalency of actin patches to endocytic sites had not yet been established. A few proteins genetically linked to endocytosis and actin were known to localize to cortical sites on the plasma membrane (e.g., Abp1 and Sac6) (Drubin et al. 1988), and immuno-electron microscopy (immuno-EM) had suggested an association between these proteins and membrane invaginations (Figure 3, A and B), leading to preliminary models for how actin might contribute to endocytosis (Mulholland et al. 1994) (Figure 3C). However at the time, there was minimal understanding of the ultrastructure, dynamics, or function of these actin structures.
An important breakthrough came in 1996, when two studies used live-cell imaging with green fluorescent protein (GFP) fusions to study actin patch dynamics (Doyle and Botstein 1996; Waddle et al. 1996). Both groups concluded that cortical actin patches move within the plane of the membrane, in some cases for long distances. It was suggested that individual patches might assemble de novo and later disassemble, but it was difficult to gather further evidence for this model, given technological limitations; e.g., the intervals in time-lapse imaging were too long to permit accurate tracking of individual actin patches. These early studies set the stage for subsequent paradigm-shifting discoveries, in which two-color live-cell imaging of pairs of cortical patch components revealed that endocytic and actin machinery colocalize but arrive at sites of endocytosis at different times (Kaksonen et al. 2003). It remained unclear though, how actin assembly was initiated at endocytic sites and how actin organization and dynamics were controlled. Efforts to connect actin with endocytosis were further confounded by unresolved differences in the molecular requirements for animal vs. yeast endocytosis, including inconsistent requirements for clathrin, the scission-promoting GTPase dynamin, and actin itself.
Genetic screens for endocytosis mutants
An obvious strength of the yeast system is the ability to conduct forward genetic screens to identify the genes involved in a cellular process. Depending on how the screen is performed, different types of genes will be identified while others will be missed (e.g., redundant gene pairs). A wide variety of genetic screens designed to identify factors important for endocytosis have been conducted over the years, with a subset of factors being found in more than one screen (Table 1). These screens have provided the platform for many subsequent reverse genetic and biochemical studies dissecting the roles of candidate genes/proteins in endocytosis and defining their physical and genetic interaction partners. Together, these approaches have yielded a rich set of information on the endocytic machinery in budding yeast (Table 2). Below, we describe several of the early studies that identified endocytic factors, which illustrates the value in performing a variety of different genetic screens to probe a single process.
The first set of endocytosis mutants came from a collection of temperature-sensitive yeast strains produced by random EMS mutagenesis. The Riezman lab initially tested this collection for strains deficient in uptake of the bulk fluid phase marker Lucifer Yellow, identifying the end1 and end2 mutants (Chvatchko et al. 1986); however, these mutations were later found to have no direct effect on the internalization process (Dulic and Riezman 1989; Hamburger et al. 1995). After switching to measuring uptake of radiolabeled α-factor (a pheromone peptide secreted by yeast), the first bona fide yeast endocytosis mutants, end3 and end4, were discovered from the same strain collection (Raths et al. 1993). Further screening identified end5, end6, and end7 (Munn et al. 1995). Interestingly, end7 is an allele of the ACT1 (actin) gene, which had been shown previously to be defective in endocytosis (Kubler and Riezman 1993). In addition, the proteins encoded by END3–END7 were later shown to localize to cortical actin patches (Benedetti et al. 1994; Vaduva et al. 1997; Wesp et al. 1997; Huh et al. 2003).
Another screen for endocytosis mutants was based on the observation that strains carrying a deletion of a vacuolar ATPase subunit such as vma2∆ grow only on low-pH media. It was speculated that the vacuoles of these mutants become acidified via the endocytic pathway (Nelson and Nelson 1990), and the Riezman lab found that combining an end mutant with vma2∆ led to inviability (i.e., synthetic lethality). By performing EMS mutagenesis in a vma2∆ background, they isolated end8 through end13 as synthetic lethal mutations (Munn and Riezman 1994). The END9 gene encodes a protein later recognized as the Arc35 subunit of the Arp2/3 complex (Machesky et al. 1994; Welch et al. 1997a), the first hint that branched actin structures are important for endocytosis (see Mechanisms of Actin Assembly and Turnover). This screen also highlighted the critical importance of membrane composition, since it identified enzymes that are required for synthesizing both sphingolipids (END8/LCB1) and the yeast cholesterol-like molecule ergosterol (END11/ERG2) (see Vesicle Scission and Uncoating).
A growing appreciation for the role of ubiquitin in endocytosis (Volland et al. 1994) led the Hicke lab to modify the original screen of random mutants defective for radiolabeled α-factor uptake. They mutagenized a strain expressing an α-factor receptor that entirely relied upon ubiquitin modification for its internalization (Terrell et al. 1998) to find ubiquitin-dependent internalization (udi) mutants. This approach yielded protein kinases Ypk1 and Tor2, which turned out to be required for both ubiquitin-dependent and -independent receptor internalization (Dehart et al. 2002; Dehart et al. 2003; Tenay et al. 2013).
Another marker for endocytosis is the fluorescent lipophilic (membrane-binding) dye FM4-64, which can be used to trace the endocytic pathway from the plasma membrane through endosomes to the vacuole (Vida and Emr 1995). EMS-mutagenized cells were subjected to a pulse/chase labeling with FM4-64, followed by flow cytometry selection of “dim” cells that had internalized the least amount of dye. This screen yielded the myosin chaperone She4/Dim1 and the EH domain scaffold protein Pan1/Dim2 (Wendland et al. 1996). A role for the lipid kinase Fab1 in late stages of the endocytic pathway was subsequently uncovered using the same approach (Shaw et al. 2003).
After the yeast genome was sequenced in 1996 (Goffeau et al. 1996), a strain collection was generated with deletions of nonessential genes, which provided the yeast community with another valuable tool for identifying endocytosis mutants. For example, a subset of these nonessential deletion strains was tested for deficiency in Lucifer Yellow accumulation, which identified the endosomal F-Box protein Rcy1, among others (Wiederkehr et al. 2000; Wiederkehr et al. 2001). Further, the Conibear lab screened several different gene deletion collections for strains with elevated plasma membrane levels of the v-SNARE protein Snc1 (Burston et al. 2009), an endocytic cargo that must be included in the membranes of secretory/endocytic vesicles for their delivery and fusion to the subsequent compartment (Gurunathan et al. 2000). This approach uncovered many of the expected genes, validating the general strategy, and identified the SH3-domain protein Ldb17 as an endocytosis protein. It also revealed quantitative endocytic defects for mutants in a number of actin binding proteins that reside on actin patches but had not previously been linked functionally to endocytosis (e.g., Aip1, Crn1, Gmf1). The Drubin lab then used a larger set of systematic gene deletions combined with temperature-sensitive alleles to screen for mutants that were resistant to the yeast killer toxin virus, some of which were resistant due to the inability to internalize the toxin (Carroll et al. 2009). While the toxin receptor remains undefined, an exciting discovery from this screen was a requirement for the yeast homolog of the AP-2 adaptor complex. Until this finding, an endocytic role for the yeast AP-2 complex had been lacking (Huang et al. 1999; Yeung et al. 1999).
Through the combined efforts of many labs, genetic screens, interaction studies, and biochemical analyses have uncovered the existence of dozens of proteins involved in endocytosis. Subsequent experimentation has shown that these components are sequentially recruited as early, middle, and late coats, leading to the formation of a colossal molecular assemblage that culminates in a burst of F-actin polymerization and the reshaping of plasma membrane into a vesicle. This multistep process is depicted in Figure 1 and described in more detail in Cargo Selection and Coat Formation, Mechanisms of Actin Assembly and Turnover, and Vesicle Scission and Uncoating.
Cargo Selection and Coat Formation
Clathrin was linked to endocytosis in animal cells in 1975 (Pearse 1975), but only in the last decade has a role for clathrin in yeast become fully accepted (Kaksonen et al. 2005; Newpher et al. 2005). Whether budding yeast had a CME pathway was initially fraught with controversy, beginning with confusion about the requirements of clathrin for viability. In one commonly used lab strain background, the clathrin heavy chain (CHC1) gene was found to be essential, while in another it was nonessential. Although the viable clathrin deletion cells were extremely sick (Lemmon and Jones 1987; Payne et al. 1987) they exhibited only partial endocytosis defects, retaining 30–50% uptake of mating pheromone (Payne et al. 1988). Later, using a chc1 temperature-sensitive allele, it became possible to acutely disrupt clathrin function, which led to reduced endocytosis within 2 min of temperature shift (Tan et al. 1993). The chronic chc1∆ and acute chc1-ts endocytosis defects were similar, indicating that clathrin might play a direct role in endocytic internalization in yeast.
Two other issues fueled suspicions about the importance of clathrin for yeast endocytosis. First, the prevailing view was that the endocytic mechanisms used in yeast were likely to differ significantly from those used in mammalian cells (Baggett and Wendland 2001), since actin had been so closely linked to yeast endocytosis while apparently playing only a minor or specialized role in mammalian cells (Gottlieb et al. 1993; Fujimoto et al. 2000). Second, the majority of clathrin in yeast cells had been localized to large, concentrated intracellular punctae that are likely Golgi associated, with no apparent localization at the cell surface (Pishvaee et al. 2000). These two observations, along with the incomplete block of internalization in clathrin mutants, led some investigators to argue against a direct role of clathrin in facilitating endocytosis in yeast.
Two key developments then led to the direct visualization of clathrin at the yeast plasma membrane. First, following on the pioneering work from the Keen and Almers labs (Gaidarov et al. 1999; Merrifield et al. 2002), Kaksonen et al. (2003) established the use of two-color live-cell imaging techniques in yeast to analyze the dynamics of pairs of components of cortical actin/endocytic patches. Through this approach, they made the landmark discovery that cortical patches form in discrete stages following a stereotypical, temporally ordered program in which endocytic components (including the clathrin heavy and light chains) are subsequently joined by actin and actin-associated proteins (Figure 1). Further, mutating the endocytic protein Sla2/End4 led to aberrantly long actin “comet tails” at endocytic sites, which allowed the measurement of both the rate of actin turnover (Kaksonen et al. 2003) and the relative positions of key actin regulatory proteins within the actin networks, i.e., at the front, at the rear, or evenly distributed (Okreglak and Drubin 2007; Gandhi et al. 2009; Michelot et al. 2013). Next, the Lemmon lab found that sla2∆ mutants and/or treatment with the actin depolymerizing drug Latrunculin-A stabilized clathrin at the plasma membrane. In contrast, they found that clathrin adaptor mutants (two epsins and two AP180 proteins) had significantly diminished clathrin accumulation at the plasma membrane (Newpher et al. 2005). These observations were supported by the use of total internal reflection fluorescence (TIRF) microscopy, which reduced the signal from bright clathrin-positive internal structures and revealed the existence of transient patches of plasma membrane-associated clathrin in wild-type cells. Importantly, these clathrin patches subsequently recruited Abp1, a well-established component of both the endocytic machinery and actin cytoskeleton (Newpher et al. 2005). Thus, clathrin was finally accepted as an important and early arriving component of the endocytic machinery in yeast, and connected CME to the actin cytoskeleton.
Overview of the stages of yeast endocytosis
A major concept revealed by the above-mentioned studies and elaborated on in studies that followed was that an endocytic site develops or matures over time and that numerous proteins are sequentially recruited to this site and assembled into a structure that collects cargo and sculpts the plasma membrane into a vesicle. This process can be broadly divided, into three operational steps: (1) an immobile phase (Figure 1, steps 1–4; coat assembly), (2) a mobile phase (Figure 1, steps 5–7; actin assembly, membrane invagination, and vesicle scission), and (3) a patch dissolution phase (Figure 1, steps 8 and 9; vesicle uncoating and fusion with endosome). Note that uncoating of certain components may also be a prerequisite to scission.
For organizational purposes, the remainder of Cargo Selection and Coat Formation focuses on the immobile phase of patch assembly. We first discuss the selection of cargo proteins and then coat assembly (Figure 1, steps 1–4), including the recruitment and functions of early, middle, and late coat complexes. As discussed in detail below, the best-characterized adaptors are members of the middle and late coat complexes. In the process of these coat layers being recruited and ultimately forming a vesicle, the protein constituents form a dense network of interactions with each other, with the membrane, and with transmembrane cargo proteins, although there are undoubtedly more interactions to be defined (Figure 2). The coats both organize cargo and generate a structural platform to transmit forces generated by F-actin polymerization for vesicle formation, as described in Mechanisms of Actin Assembly and Turnover and Vesicle Scission and Uncoating.
Cargo sorting signals and adaptors
Endocytic adaptors recognize cytoplasmic sorting signals in cargo proteins and couple the cargo to the coat. There are generally two types of sorting signals: short peptide motifs and covalently attached ubiquitin molecules (Reider and Wendland 2011). Different adaptors recognize different sorting signals on transmembrane proteins at the plasma membrane. Key open questions include uncovering the code that matches each adaptor with each cargo and learning whether particular classes of cargo are targeted by adaptors in response to specific environmental changes.
Ubiquitin as a sorting signal:
Ubiquitin (Ub) is a conserved, 76 amino acid protein that can be covalently linked to substrate proteins via a cascade of three enzymes: E1 (Ub-activating), E2 (Ub-conjugating), and E3 (Ub-ligase) (Varshavsky 2012). Commonly, Ub is linked to a lysine residue in the substrate protein, and since Ub itself has lysines on its surface, it can act as a target for Ub-ligases to produce poly-Ub chains. In mammalian cells, Ub modifications are important for intracellular sorting events at endosomal compartments (Erpapazoglou et al. 2012). There were hints that Ub might contribute to endocytic regulation in yeast starting in the mid-1990s; high-copy Ub was found to suppress the lethality associated with clathrin deletion (Nelson and Lemmon 1993), and some plasma membrane cargoes were found to be ubiquitylated prior to their internalization (Kolling and Hollenberg 1994; Hein et al. 1995; Hicke and Riezman 1996; Roth and Davis 1996). A key role of Ub as a signal for endocytic internalization was shown by the translational fusions of Ub to a mating pheromone receptor (Terrell et al. 1998). We now know that Ub-labeled endocytic cargo can either be mono-ubiquitylated (note that more than one lysine can be modified, generating a multi-ubiquitylated protein) or bear K63-linked poly-Ub chains. The major enzyme responsible for covalent attachment of Ub to endocytic transmembrane cargo proteins is the E3 ligase Rsp5 (Springael et al. 1999). Rsp5 substrate selection is influenced by protein kinases such as Npr1 (Kaouass et al. 1998; De Craene et al. 2001; Macgurn et al. 2011; Merhi and Andre 2012), raising the question of how the kinase activities are controlled.
α-Arrestins bridge Rsp5 and cargo substrates:
Rsp5 contains three WW domains that promote binding to [L/P]PxY motifs in its substrates (Kasanov et al. 2001); however, many Rsp5-dependent substrates, including transmembrane proteins, lack [L/P]PxY motifs. Recent work has uncovered an important role for the [L/P]PxY motif-containing α-arrestin family in bridging Rsp5 to specific transmembrane cargo destined for covalent Ub modification and internalization. This family consists of 13 known members, many with partially redundant functions; additionally, many cargo proteins are able to functionally interact with multiple α-arrestins (Lin et al. 2008; Nikko and Pelham 2009). The precise mechanisms by which cells regulate binding of transmembrane cargo to α-arrestins are unclear, but may be dictated by physiological or environmental conditions (Nikko and Pelham 2009; Macgurn et al. 2011; Merhi and Andre 2012). Another recent finding involves the action of the N-terminal “LID” domain found in permeases, a transmembrane cargo family; this LID domain undergoes a conformational change when the permease is nonfunctional or misfolded, triggering an Rsp5-dependent (but α-arrestin independent) Ub modification (Keener and Babst 2013). It will be interesting to learn more about α-arrestin-dependent and -independent routes that govern plasma membrane protein turnover.
Ub-binding endocytic adaptors:
The first ubiquitin interaction motif (UIM) identified found was a conserved α-helical motif in the Rpn10/S5a subunit of the proteasome S19 domain (Hofmann and Falquet 2001). The realization that a UIM could predict Ub-binding activity stimulated the study of a number of endocytic proteins with UIMs, in particular the epsin proteins. The two epsin proteins in S. cerevisiae, Ent1 and Ent2, are members of the middle coat. Initial studies suggested that the epsins might act as endocytic adaptors for binding to covalent Ub modifications on cargo (Aguilar et al. 2003). More recent work suggests that these and other Ub-binding proteins (e.g., Ede1 and its UBA domain) may instead utilize their Ub-binding regions for self-regulation, similar to what has been proposed for their mammalian counterparts (Dores et al. 2010). Another Ub-binding adaptor protein is the late coat protein Sla1, which uses one of its SH3 domains to bind Ub via an unconventional interaction (Stamenova et al. 2007). Interestingly, Sla1 and its mammalian homolog CIN85 also bind to Ub E3 ligases, a requirement for the internalization of certain receptors (Szymkiewicz et al. 2002; Tonikian et al. 2009).
Peptide-based sorting motifs and their adaptors:
The concept of peptide-based sorting motifs originally emerged from studies of familial hypercholesterolemia, in which it was discovered that the Tyr-x-x-hydrophobic signal in the cytosolic tail of the LDL (low density lipoprotein) receptor directs internalization of LDL particles (Anderson et al. 1977). Since then, numerous other peptide-based sorting motifs have been recognized in the tails of many animal cell cargo proteins, mediating selection and sorting at various organelles along the secretory and endocytic pathways.
The only endocytic peptide sorting motif characterized in yeast to date is the NPF(x)1,2D sorting motif (Tan et al. 1996). NPF tripeptide motifs are hereafter labeled as NPFmotif, to distinguish them from the NPF acronym used later (see Mechanisms of Actin Assembly and Turnover) in reference to actin nucleation promoting factors (NPFs). Although most NPFmotifs are bound by EH domains, typically the NPF(x)1,2D sorting motif is not (Paoluzi et al. 1998; Howard et al. 2002). Instead, the NPF(x)1,2D motif binds the adaptor protein Sla1, via its SH3-like SHD1 (Sla1 homology domain 1) (Howard et al. 2002; Mahadev et al. 2007). One possible reason for this distinction may be the influence of other Sla1 binding partners, such as membrane or cargo (via Ub), but more work is needed to resolve this. Sla1-interacting transmembrane cargo bearing the NPF(x)1,2D motif (or a variant) include the mating pheromone receptors Ste2 and Ste3, the cell-wall stress sensor Wsc1, and the phospholipid flipases Drs2 and Dnf1 (Howard et al. 2002; Liu et al. 2007; Piao et al. 2007).
In animal cells, the C-terminal extension of the μ-subunit of the heterotetrameric AP complexes (and related domains), known as the μHD domain, has well-known peptide-sorting motif-binding activities (Owen and Evans 1998; Jung et al. 2007; Michelsen et al. 2007). The yeast early coat protein Syp1 contains a μHD domain, suggesting a role for Syp1 as an adaptor (Reider et al. 2009). It is interesting to define the sorting motif recognized by the Syp1–μHD and to learn whether other yeast adaptors similarly use peptide sorting motifs. In addition, it is important to continue the search for new adaptors.
Initiation of endocytic sites
Endocytic proteins in the early coat are the first to arrive at a nascent endocytic site during the initiation phase (Figure 1, step 2), and include clathrin, AP-2, Ede1, Syp1, and Pal1 (Kaksonen et al. 2005; Boettner et al. 2009; Carroll et al. 2009; Reider et al. 2009; Carroll et al. 2012). There are many questions about the early coat proteins, including what their functions are (both individually and as a group), what triggers their initial recruitment to the membrane, and why the majority of endocytic sites occur at/near sites of polarized growth. Another question is how yeast clathrin is recruited to these sites, since none of the early coat proteins are known to bind clathrin (Figure 2), with the possible exception of YAP180 (whose arrival time is debated in the literature; see below). Interestingly, among all the stages of endocytosis, the initiation phase is the most variable in length (Newpher and Lemmon 2006; Toret et al. 2008; Boettner et al. 2009; Carroll et al. 2012), raising questions about what leads to this variability and what triggers or checkpoints govern progression to subsequent recruitment of middle coat proteins.
Early coat proteins:
Two of the earliest-arriving coat proteins, Ede1 and Syp1, directly bind each other (Boettner et al. 2009; Reider et al. 2009; Stimpson et al. 2009). Ede1 most closely resembles the mammalian endocytic protein Eps15, which is classified as a combination scaffold/ubiquitin adaptor (Gagny et al. 2000; Polo et al. 2003). The N-terminal region contains three EH domains (which bind NPFmotifs), and the C-terminal region includes an ubiquitin-binding UBA domain. The mammalian proteins FCHO1 and FCHO2 are homologs of Syp1 and have been implicated in stabilizing nascent clathrin-coated pit structures (Henne et al. 2010; Cocucci et al. 2012). Members of this family have a domain layout consisting of an N-terminal F-BAR domain (an anti-parallel homodimerizing membrane-binding domain), a central unstructured region that contains a proline-rich domain, and a C-terminal cargo-binding μHD (see above; Owen and Evans 1998; Reider et al. 2009). While the F-BAR domain binds to membranes with a particular curvature (Heath and Insall 2008), the unstructured central region contains a high-affinity membrane-binding activity that is sufficient to mediate strong membrane association. This region of Syp1 also directly binds and inhibits the actin polymerization functions of Las17 ((Boettner et al. 2009; Feliciano and Di Pietro 2012); see Mechanisms of Actin Assembly and Turnover). Although ede1∆ or syp1∆ cells have relatively mild endocytic phenotypes, these defects are consistent with the observed roles in mammalian cells, such as influencing the localization of clathrin and reducing the number and stability of endocytic patches (Kaksonen et al. 2005; Reider et al. 2009; Stimpson et al. 2009; Carroll et al. 2012).
The early coat protein Pal1 is another member of an Ede1-containing complex (Gavin et al. 2002; Carroll et al. 2012). Pal1 appears to be specific to fungi and has no characterized domains that hint at its functions, although it does have two copies of the NPFmotif. The S. pombe homolog of Pal1 is implicated in regulating cell polarity (Ge et al. 2005), which is interesting since endocytic sites in S. cerevisiae initiate in a polarized manner (Adams and Pringle 1984). The S. cerevisiae genome encodes a paralog of Pal1, called Yhr097c, but it seems unlikely to contribute to endocytosis given its localization to intracellular sites and its relative lack of genetic or physical interactions with other endocytic machinery (Saccharomyces Genome Database, http://www.yeastgenome.org).
Yeast AP-2 complex is also an early coat component. It is homologous to the mammalian AP-2 complex, except that the yeast complex does not bind clathrin (Yeung et al. 1999). As mentioned above, yeast AP-2 has a role in killer toxin internalization. In addition, it was recently shown to be involved in the maintenance of cell-wall integrity by acting as the endocytic cargo adaptor for the cell-wall stress-sensor Mid2 (Chapa-Y-Lazo et al. 2014). Beyond these roles, little else about the function and regulation of AP-2 in yeast is known. Thus, many questions remain open about its recruitment to the early coat and/or the plasma membrane and how it interacts with other early coat proteins.
Early coat protein dynamics are variable:
While the temporal behaviors of most endocytic protein “modules” (Figure 1) are invariant in wild-type cells (Kaksonen et al. 2005), the early coat is an exception, ranging in lifetime from 30 to 180 sec (Newpher and Lemmon 2006; Toret et al. 2008; Boettner et al. 2009; Carroll et al. 2012). This variability may reflect a putative checkpoint that monitors the packaging of cargo at an endocytic site (Loerke et al. 2009; Carroll et al. 2012) and restricts progress until cargo is suitably concentrated or organized. Recently, it was found that the regions of the cell with the highest concentrations of newly delivered exocytic proteins (some of which are undoubtedly endocytic cargo) colocalize with the most rapidly maturing early coat patches (Layton et al. 2011), consistent with the checkpoint model. However, the major cargo-binding components of the endocytic patch are not recruited until later steps (see below). Whatever the precise cause of temporal variability, it is clear that some trigger causes a transition within the early coat proteins and/or the underlying membrane that allows the next wave of proteins to join the endocytic site.
Early coat proteins appear to have differing roles in regulating later-arriving endocytic proteins. Both ede1∆ and chc1∆ cells display shorter residence times of the late coat protein Sla1 (see below) and a reduced number of Sla1-containing patches at steady state (Kaksonen et al. 2005; Newpher and Lemmon 2006). Intriguingly, these effects on late coat behavior are apparently specific to chc1 and ede1 mutants and are not seen in pal1, syp1, or AP-2 mutants (Carroll et al. 2009; Carroll et al. 2012). Further analysis may shed light on how clathrin and Ede1 influence later-arriving endocytic proteins and/or other early coat components and may uncover a link to the putative cargo checkpoint.
Another feature of the early coat is the differential spatial behavior of its constituents. Clathrin, AP-2, and Pal1 all move inward with the coat during the mobile phase, while Ede1 and Syp1 remain at the plasma membrane (Stimpson et al. 2009). This is consistent with the activity of Syp1, which inhibits the polymerization of actin (see below); perhaps Ede1 also negatively regulates an activity required for the later stages and must disperse to promote the mobile phase.
Maturation of endocytic sites
“Maturation” refers to the events that occur between the end of the variable early coat initiation phase and the beginning of the mobile invagination/scission phase and typically lasts for about 30 sec. During this interval, the coat transitions to a more stereotypical behavior and prepares for assembly of the actin network. Initially, characterization of the temporal and spatial dynamics of endocytic proteins had not included the early coat proteins and was focused on the immobile coat proteins that became mobile when F-actin appeared (Kaksonen et al. 2003). However, subsequent studies with higher-resolution time-lapse analyses of many more proteins revealed a need to further classify endocytic components into middle coat vs. late coat proteins (Chi et al. 2012). This more refined classification should facilitate the identification of which proteins influence the transitions from initiation to maturation to scission. The middle coat proteins represent major links to clathrin and cargo (Figure 2), while the late coat proteins are regulators of actin dynamics (see Mechanisms of Actin Assembly and Turnover). As discussed in more detail below, there is some uncertainty as to whether the membrane curvature/invagination stage precedes or coincides with the appearance of actin.
Middle coat proteins:
The first middle coat proteins to join the early coat patch are likely to regulate the transition between the end of the early coat-based initiation phase and the next step in endocytosis. Several observations point to Sla2 playing a role in this transition. First, Sla2 physically interacts with three early coat proteins (Ede1, Syp1, and Pal1) (Gavin et al. 2002). Second, Sla2 arrives ∼6 sec before the late coat protein Sla1 (Newpher and Lemmon 2006). Third, an interaction between Sla2 and clathrin light chain in turn affects the timing of Sla1 recruitment (Newpher and Lemmon 2006; Boettner et al. 2011), suggesting that the early coat together with Sla2 influences later stages of endocytosis.
Sla2 is the yeast homolog of HIP1R (Huntingtin interacting protein-1 related). Sla2 and HIP1R are both homodimeric proteins and have N-terminal membrane-binding ANTH (AP180 N-terminal homology) domains, central coiled-coil regions that bind clathrin light chain, and C-terminal F-actin-binding THATCH (talin-HIP1/R/Sla2p actin-tethering C-terminal homology) domains (Wesp et al. 1997; Yang et al. 1997; Henry et al. 2002). In both yeast and mammals, Sla2/HIP1R has been implicated in linking the endocytic coat to F-actin and may direct actin polymerization and force generation to facilitate membrane invagination. In the absence of Sla2/HIP1R, uncontrolled polymerization of F-actin occurs without associated vesicle formation, resulting in actin comet tail structures (Kaksonen et al. 2003). This is now called an uncoupling phenotype, where actin polymerization appears to be unlinked from membrane invagination (Brady et al. 2010; Skruzny et al. 2012). However, the comet tails in sla2∆ cells are likely associated with at least small membrane invaginations, as suggested by immuno-EM studies and by the observation that FM4-64 accumulates in plasma membrane-associated punctae in these cells (Stefan et al. 2002; Idrissi et al. 2012). Intriguingly, clathrin mutants also exhibit actin comet tails; they are distinct in shape and size from those found in sla2∆ cells, but this still suggests that clathrin may also be involved in the coupling of the coat to actin, perhaps via the Sla2/Clc1 interaction described above (Newpher and Lemmon 2006).
The yeast epsins Ent1 and Ent2 have been defined as middle coat proteins based on both the timing of their arrival at patches and their functional links to Sla2 (Toret et al. 2008; Skruzny et al. 2012). Ent1 and Ent2 are an essential pair of clathrin adaptor proteins whose epsin N-terminal homology (ENTH) domains bind PtdIns(4,5)P2 and play a crucial role in signaling to the polarity-regulating GTPase Cdc42 (Wendland et al. 1999; Aguilar et al. 2006). Ent1 and Ent2 each contain two ubiquitin-interaction motifs (UIMs) adjacent to their ENTH domains; further, their intrinsically disordered C-terminal domains harbor an NPFmotif and a variant clathrin box motif that binds the terminal domain of clathrin (Wendland et al. 1999; Kalthoff et al. 2002; Wendland 2002). The epsin C-terminal domain also binds actin (Skruzny et al. 2012).
Recent evidence has implicated Ent1 and Ent2 in the transmission of force from actin polymerization to the plasma membrane, in complex with Sla2. First, the Sla2 THATCH domain was found to be required for endocytosis in ent1 ent2 mutant cells (Baggett et al. 2003). Second, the lipid binding ANTH domain of Sla2 heterodimerizes with the lipid-binding ENTH domain of Ent1 or Ent2 and cooperatively binds the plasma membrane (Skruzny et al. 2012).
Two additional coat proteins are Yap1801 and Yap1802, yeast homologs of the clathrin assembly lymphoid myeloid leukemia protein (CALM), a ubiquitously expressed clathrin adaptor (Dreyling et al. 1996; Chi et al. 2012). There is some disagreement in the literature about the precise temporal recruitment of the Yap180 proteins, as different groups observed them in either the early or middle coat, possibly influenced by the different positions of the GFP tags (Carroll et al. 2012; Chi et al. 2012). Yap1801/2 in yeast and CALM homologs in other organisms have conserved functions in endocytosis and in recycling of the v-SNARE Snc1/VAMP (Maritzen et al. 2012). Like Sla2, the Yap1801/2 proteins have N-terminal lipid-binding ANTH domains, and like the epsins they have intrinsically disordered C termini; Yap1801/2 and the epsins also have common binding partners (clathrin and EH domain scaffolds) (Wendland and Emr 1998). Further, yeast cells lacking all four of these related proteins (Ent1, Ent2, Yap1801, and Yap1802) exhibit temperature-sensitive growth and endocytic defects (Maldonado-Baez et al. 2008).
Late coat proteins:
Another cohort of endocytic proteins defines the late coat (Kaksonen et al. 2005; Chi et al. 2012). Many late coat proteins associate with each other, with early and middle coat proteins, and with actin regulatory factors recruited in the final invagination/scission stage (Figure 2). As described below, Sla1 is currently thought to shepherd the transition from the middle to late coat stages.
Sla1 is a critical late coat protein that was among the first clear links discovered between the endocytic and actin machineries (Warren et al. 2002). Sla1 has numerous binding partners and regulatory interactions and the capacity to both positively and negatively regulate endocytosis. Its domain layout consists of three SH3 domains, two distinct conserved Sla1 homology domains (SHD1 and SHD2), a clathrin-binding motif, and a C-terminal region that is heavily phosphorylated by actin-regulating kinase (ARK) family kinases (see Vesicle Scission and Uncoating). Sla1 is subject to autoinhibition, as its SHD2 domain binds to and inhibits the clathrin-binding motif (Di Pietro et al. 2010). Sla1 forms a complex with two EH domain-containing late coat proteins, Pan1 and End3 (Tang et al. 1997; Tang et al. 2000). Recently, the specific domains in Pan1 and End3 mediating their interaction were identified, and it was shown that this interaction is required for Pan1 phosphoregulation (Whitworth et al. 2014). This Sla1–Pan1–End3 trio may be the yeast functional counterpart to mammalian endocytic scaffold/signaling protein Intersectin, which has EH and SH3 domains in a single polypeptide (Yamabhai et al. 1998).
Studies have shown that once Sla1 is recruited, an endocytic patch is highly likely to assemble F-actin and undergo internalization (Newpher and Lemmon 2006). Thus, the mechanism of Sla1 recruitment appears to be linked to the mechanism that deems a patch ready to transition from the immobile to the mobile phase. What underlies this key transition step remains unknown, but is likely affected by phosphoregulation (see below). It is possible that Sla1 monitors patch maturation by “sensing” the status or concentration of clathrin. As noted above, Sla1 residence in patches is reduced in chc1∆ cells (Kaksonen et al. 2005), and cells expressing a sla1 mutant that cannot bind clathrin are defective for internalization of the Sla1-dependent cargo protein Wsc1 and have significantly longer lived patches (Di Pietro et al. 2010). However, further work is needed to fully understand this mechanism.
Pan1 is another critically important late coat protein; surprisingly, it is the only component of the endocytic coat that is essential for viability (in addition to the clathrin heavy chain in certain strain backgrounds mentioned above) (Sachs and Deardorff 1992; Wendland et al. 1996). Whether the essential function of Pan1 is tied to its role in endocytosis has not yet been determined, but given its subcellular localization to cortical patches, the phenotypes of pan1 mutant cells, and numerous physical and genetic interactions (Miliaras and Wendland 2004), it is clear that it plays a central, and perhaps essential, role in endocytosis.
Pan1 was originally misidentified as a poly(A)-binding protein-dependent poly(A) nuclease (hence its name, Pan1), but Pan2 was later shown to harbor this nuclease activity (Sachs and Deardorff 1992; Boeck et al. 1996). After realizing that Pan1 in fact is an endocytic protein, the clathrin-binding adaptor proteins Yap1801/2 and Ent1/2 were identified as binding partners of Pan1’s conserved EH domains (Wendland and Emr 1998; Wendland et al. 1999). The Cai lab independently connected Pan1 to endocytosis through suppressor screens, which also identified the late coat proteins Sla1 and End3 and the regulatory protein kinase Prk1 (a member of the ARK kinase family) (Tang et al. 1997, 2000). The Drubin lab later showed that Pan1 binds the Arp2/3 complex and can stimulate its actin nucleation activity, thereby linking Pan1 to actin regulation (Duncan et al. 2001). Further, Sla2 interactions with Pan1 were shown to inhibit its effects on Arp2/3 complex, providing a potentially important regulatory mechanism (Toshima et al. 2007). Pan1 also binds to the type I myosins Myo3/5 and facilitates their ability to stimulate Arp2/3 complex (Barker et al. 2007). These many binding partners and activities raise a key question about Pan1, which extends more generally to other coat proteins, which is how do these proteins interact with their binding partners in the correct temporal sequence when the partners are often recruited at distinct stages of the endocytic process?
A final late coat protein, Las17, arrives coincidentally with Pan1. Since Las17 is one of the most potent activators of Arp2/3 complex-mediated actin polymerization, it is discussed below (Mechanisms of Actin Assembly and Turnover).
Mechanisms of Actin Assembly and Turnover
In this section we focus on events that occur at endocytic sites after the recruitment of the early, middle, and late coats (above) and that produce a short and critically required burst of actin assembly that drives vesicle scission and internalization (Kaksonen et al. 2003, 2005) (Figure 1 and Figure 3). Although this phase of endocytosis lasts only 15–20 sec, it is the major force-generating step in the endocytic process and results in all of the extreme membrane remodeling events that accompany endocytic internalization. Patches also reach their peak of molecular complexity at this phase, since most earlier components remain present and >30 new proteins arrive (Table 2).
Over the past 17 years, there have been major advances in defining the activities, mechanisms, and structures of actin-associated components, as well as their physiological interacting partners, dynamics, and cellular roles. Moreover, from quantitative live-cell imaging studies we know the timing of arrival of most actin-associated protein components, their molar ratios in cells, and even the number of molecules per endocytic site for many of these proteins. Finally, EM studies have defined the shape and dimensions of endocytic invaginations and surrounding actin networks (Figure 3, G and H). With this information, it has been possible to begin answering central mechanistic questions about this actin-based phase of endocytosis, including: How is the timing of actin assembly regulated at endocytic sites, and what mechanisms initiate actin polymerization? How are the size, shape, and geometry of the actin networks determined? How is a particular set of actin-associated proteins recruited to the actin filaments in patches, while a different set is recruited to the other major actin structures in yeast: actin cables and the cytokinetic ring (Figure 4)? Finally, how does the actin patch network provide directional force to drive vesicle invagination and scission?
The working model we arrive at (described below) helps to explain how >30 different actin regulatory proteins arrive at patches within 5–7 sec of each other and work in concert to initiate the rapid formation of an actin network, stimulate its dynamic turnover and remodeling, and ultimately dismantle the network and recycle its components (Figure 4). In many ways, this process can be compared to Listeria actin-based motility, which requires both focused nucleation at the bacterial surface and rapid turnover of the network (Tilney and Portnoy 1989). Similarly, latex beads coated with actin nucleators found at yeast endocytic sites (see below), when mixed with yeast-cell extracts, reconstitutes in vitro bead motility (Michelot et al. 2010). Force generation by actin networks in both motility and endocytosis can be explained by the Brownian ratchet and elastic Brownian ratchet models, in which actin polymerization produces work by polymerizing to fill small gaps between the actin network and membrane created by Brownian motion (Kaksonen et al. 2006). In these models, actin polymerization must be focused at the membrane, consistent with nucleation and elongation machinery being typically found at these sites. Capping proteins are additionally required to focus polymerization at new nucleation sites and to maintain network architecture, and filament branching and crosslinking factors are necessary to hold a network together to enable force production. The precise organization and orientation of actin filaments in yeast actin patches is still being worked out, but most evidence suggests that the barbed ends face the cell cortex, at least at early stages of membrane invagination (details below).
In the subsections below, we describe in detail the molecular machinery and mechanisms underlying each major step in actin regulation that occurs during yeast endocytosis: (A) actin nucleation, (B) actin-filament capping, crosslinking, and stabilization, and (C) actin-filament disassembly and turnover (Figure 4). Genetic analysis has shown that each step is crucial for endocytosis, presumably because the rapid assembly and turnover of the actin networks provides: (i) directional force to drive plasma membrane tubulation and vesiculation and (ii) plasticity necessary for the actin network surrounding the membrane invagination to be dynamically resculpted as the geometry of the membrane changes (Figure 1, G and H).
Actin is a ubiquitous, abundant, and conserved 42-kDa protein that binds ATP and exists in either a globular (monomeric) form called G-actin or a filamentous (polymeric) form called F-actin. Actin filaments are polarized, helical, two-stranded polymers with a fast-growing “barbed” end and a slow-growing “pointed” end, named for their appearances in electron micrographs after decoration with myosin S1 fragment. The critical concentration for assembly of yeast and animal actins is ∼0.15 µM, yet their cytosols maintain micromolar levels of actin monomers; thus cellular conditions strongly favor actin assembly over disassembly. What limits the growth of actin filaments in yeast and animal cells is the production and/or availability of “free” (uncapped) barbed ends of filaments onto which actin subunits can be rapidly added. Because free ends are limiting, a reservoir of ATP-actin monomers builds up in cells, poised for growth once free ends become available. Profilin is a small, abundant cellular protein that preferentially binds ATP-actin monomers and blocks their addition onto pointed ends but freely allows addition onto barbed ends (Tilney et al. 1983; Pollard and Cooper 1984; Haarer et al. 1990; Haarer et al. 1993; Eads et al. 1998). Thus, actin polymerization in vivo is restricted to barbed ends, and the generation of free barbed ends is a key regulatory step in actin network formation.
In principle, there are three distinct mechanisms by which free barbed ends can be generated. Existing capped filaments can be uncapped, existing filaments can be severed to produce new ends, and new filaments can be formed de novo from monomers. Although there is some evidence for uncapping and severing contributing to the growth of actin networks in animal cells (Condeelis 2001), rich evidence demonstrates that the primary mechanism for stimulating actin network growth is regulated nucleation (Pollard and Borisy 2003; Pollard and Cooper 2009).
While actin monomers are rapidly added to existing filaments barbed ends (11.6 subunits sec-1 µM-1; Pollard 1986), they do not readily self-nucleate to form new filaments in the absence of other factors. This is because polymerization intermediates, actin dimers and trimers, are extremely short lived and rapidly dissociate back into monomers (Sept and McCammon 2001). In addition, profilin further suppresses nucleation by blocking longitudinal (long-pitch) actin–actin contacts (Schutt et al. 1993). In yeast and other organisms, these kinetic barriers to nucleation are overcome by activation of specialized actin regulators, called nucleators, which structurally mimic a filament end, stabilize polymerization intermediates, or recruit and organize monomers into filament seeds (Chesarone and Goode 2009; Campellone and Welch 2010).
Two distinct sets of actin assembly-promoting machinery have been identified in yeast, the Arp2/3 complex and formins (Figure 4). Endocytic actin patches are nucleated by the combined activities of Arp2/3 complex and its cofactors, called nucleation-promoting factors (NPFs, unrelated to the NPFmotif discussed in Cargo Selection and Coat Formation). Actin cables and cytokinetic rings are nucleated instead by formins (Bni1 and Bnr1) working in combination with a different set of cofactors, which includes Bud6, profilin, and tropomyosin. Yeast actin structures (patches, cables, and rings) each have a distinct shape, function, architecture, and molecular composition (Moseley and Goode 2006). How these differences are programmed is not yet understood, but may stem in part from which nucleator is used to initiate the actin network (Michelot and Drubin 2011). Further, the relative amount of actin distributed between different types of actin networks (e.g., patches and cables) is controlled by homeostatic mechanisms, in which the nucleators in these different structures compete for a limiting pool of G-actin in the cytosol (Gao and Bretscher 2008; Burke et al. 2014). This balance can also be shifted by any of the complex sets of stimulatory and inhibitory interactions that control the activity states of the nucleator systems (see below).
The Arp2/3 complex:
The Arp2/3 complex (∼200 kDa), which was first isolated from Acanthamoeba cell extracts (Machesky et al. 1994), consists of seven subunits and is conserved across distant species of animals, plants, and fungi (Welch et al. 1997b, 1998). It was also the first bona fide actin nucleator identified (Welch et al. 1997b; Mullins et al. 1998) and remains the only one that nucleates branched filaments, polymerizing new “daughter” filaments from the sides of existing “mother” filaments at a 70° angle (Mullins et al. 1998). Thus, the Arp2/3 complex forms branched networks, such as those found at the leading edge of motile cells (Svitkina and Borisy 1999) and at endocytic sites in yeast and mammalian cells (Young et al. 2004; Rodal et al. 2005a; Collins et al. 2011). Given its two actin-related protein (Arp) subunits (Figure 5A), there was early speculation that the complex might directly nucleate actin polymerization (Mullins et al. 1996), and shortly thereafter seminal work from the Mitchison and Pollard labs demonstrated that this is the case (Welch et al. 1997b, 1998; Machesky et al. 1999). Importantly though, the Arp2/3 complex alone has minimal activity, and strong nucleation requires binding of an NPF. Four NPFs have been identified in yeast (see below), with the Wiskott–Aldrich syndrome protein (WASP), called Las17 in yeast, being the best characterized. Las17/WASP contains a C-terminal VCA (or WCA) region (detailed below), which is sufficient to bind Arp2/3 complex and actin monomers and stimulate actin nucleation in vitro.
A critical advance in our understanding of the Arp2/3 complex mechanism came with publication of its crystal structure, solved for bovine Arp2/3 (Robinson et al. 2001) (Figure 5A). This provided a wealth of information to guide subsequent genetic and biochemical dissections of individual subunits defining Arp2/3 function, mechanism, and regulation in yeast (Beltzner and Pollard 2004; D’agostino and Goode 2005; Martin et al. 2005; Daugherty and Goode 2008; Balcer et al. 2010; Goley et al. 2010). The structure of the complex also revealed that the Arp2 and Arp3 subunits were well separated from each other, suggesting that an activation step might be required to bring Arp2 and Arp3 into register for nucleation. This hypothesis was confirmed both by EM and FRET studies, which showed that WASP triggers a conformational change in yeast or human Arp2/3 complex, shifting it from an “open” (inactive) state to a “closed” (primed for nucleation) state (Figure 5B) (Goley et al. 2004; Rodal et al. 2005b). Further insights came from electron micrographs and 3D reconstructions of yeast Arp2/3 complex at filament branch sites, which revealed that additional conformational rearrangements in the complex occur upon mother filament binding (Rouiller et al. 2008). Moreover, addition of preformed filaments was shown to catalyze nucleation by VCA and bovine Arp2/3 complex (Higgs et al. 1999). These results suggest that robust nucleation by Arp2/3 complex requires interactions with both an NPF and a mother filament.
Studies comparing yeast and mammalian Arp2/3 complex activities have revealed some differences that may be relevant, including partial nucleation activity by yeast Arp2/3 complex in the absence of NPFs, i.e., weak activation by F-actin alone (Winter et al. 1999; Goode et al. 2001; Rodal et al. 2005b; Wen and Rubenstein 2005). However, genetic studies show that NPFs are essential for cell viability and endocytosis (Evangelista et al. 2000; Lechler et al. 2000); thus, nucleation by Arp2/3 without NPFs is not sufficient in vivo. Still, this low-level NPF-independent nucleation activity of yeast Arp2/3 complex could be important during endocytosis; e.g., NPFs may control network initiation as well as ongoing polymerization at the membrane, while NPF-independent nucleation may contribute to reinforcing the network away from the membrane.
One of the ultimate goals in studying Arp2/3 complex mechanism is to define the precise order and timing of each step in the nucleation pathway. This is extremely challenging however, because it involves a complex set of interactions involving Arp2/3 complex, NPFs, F-actin, and G-actin. Some of the individual steps in this pathway were elegantly defined by measuring the rate constants of specific interactions involved, allowing investigators to build models of the pathway (Beltzner and Pollard 2008). However, key steps in such models required speculation where direct measurements were not available, which left gaps in our understanding. More recently, multi-wavelength single-molecule analysis of fluorescently labeled actin, yeast Arp2/3 complex, and dimerized mammalian WASP–VCA (di-VCA) molecules has allowed direct visualization of the events in real time, making it possible to quantitatively define each step in the pathway (Figure 5C) (Smith et al. 2013a,b). These studies demonstrated that di-VCA first binds Arp2/3 complex in solution and then diVCA-bound Arp2/3 complex associates with the side of a mother filament, and finally a daughter filament begins to polymerize, but only after di-VCA dissociates from Arp2/3 complex (Figure 5D).
Importantly, these studies identified several highly inefficient steps in the nucleation pathway, leaving generous room for positive regulation by additional cellular factors. For example, in vitro the great majority of complexes that bind the mother filament rapidly dissociate without ever nucleating a daughter filament, so factors that stabilize Arp2/3 complex on mother filament sides could dramatically enhance nucleation. Single-molecule studies now offer a powerful strategy by which to define the precise spatial and temporal effects of each NPF, alone and together, on nucleation and branching (Figure 5C).
NPFs directly associate with Arp2/3 complex and actin and serve as cofactors in stimulating nucleation. As mentioned above, there are four known NPFs in yeast: (1) Las17/WASP, (2) myosin I (Myo3 or Myo5) in a complex with Vrp1/WIP, (3) Pan1/Intersectin, and (4) Abp1 (Winter et al. 1999; Evangelista et al. 2000; Lechler et al. 2000; Duncan et al. 2001; Goode et al. 2001) (Figure 6). Each of these NPFs localizes to actin patches, binds to actin and to Arp2/3 complex, stimulates Arp2/3 nucleation activity to some degree in vitro, and exhibits genetic interactions with Arp2/3 complex and/or other NPFs. Las17/WASP and myosin I (with Vrp1) are categorized as class I NPFs because they bind actin monomers and have strong NPF effects (Sun et al. 2006). Pan1 and Abp1 are categorized as class II NPFs because they bind to F-actin and have comparably weak NPF effects (Figure 6C). Consistent with this classification, Arp2/3-inactivating mutations in Las17 combined with mutations in either myosin I or Vrp1 are synthetic lethal, demonstrating that class I NPF function is essential (Evangelista et al. 2000; Lechler et al. 2000). The roles of Pan1 and Abp1 in regulating Arp2/3 complex during endocytosis are less well understood (see below).
Las17 (also known as Bee1) is the yeast homolog of WASP and was first identified as a factor essential for reconstituted actin patch assembly in a permeabilized cell assay (Li 1997) and then as a multicopy suppressor of vrp1 mutants (Naqvi et al. 1998). Las17 contains an N-terminal WASP-homology 1 (WH1)/EVH1 domain that binds to Vrp1/WIP, followed by a central proline-rich region that binds to a number of SH3 domain-containing proteins (e.g., Sla1, Bbc1, Bzz1, Myo3, Myo5, Lsb1, Lsb2, Ysc84, Sho1, and Rvs167), and finally a C-terminal VCA region (Figure 6). The ∼10-kDa VCA region consists of a WASP-homology 2 (WH2 or V) domain that binds G-actin, a short “connecting” (C) domain that binds Arp2/3 complex, and an “acidic” (A) domain that binds Arp2/3 complex. Biochemical analyses of yeast and animal WASP homologs have demonstrated two critical mechanistic roles for VCA during nucleation: (i) it induces conformational changes in Arp2/3 complex that promote nucleation, and (ii) it delivers the first actin monomer of the daughter filament via its WH2 (or V) domain (Higgs et al. 1999; Rohatgi et al. 1999; Rodal et al. 2003) (Figure 5D). Early models of nucleation proposed binding of a single VCA to a single site on Arp2/3 complex. However, it was later shown that dimerization of human VCA elevates NPF effects substantially and increases VCA affinity for Arp2/3 complex by >100-fold compared to monomeric VCA (Padrick et al. 2008). More recently, it was found that there are two separate VCA-binding sites on bovine Arp2/3 complex (Padrick et al. 2011; Ti et al. 2011). Together, these studies suggest that maximal nucleation is achieved by Arp2/3 complex binding two VCAs simultaneously. Another implication of these findings is that the activity of WASP, and possibly other NPFs, can be enhanced by ligand-induced dimerization, and indeed, specific binding partners of Las17 (e.g., Bzz1, which has a dimeric F-BAR domain, Figure 6C) may function in this capacity.
While the VCA region is critical for yeast Las17 NPF effects, even dimerized it has substantially weaker NPF effects compared to full-length Las17 (Rodal et al. 2003). These observations suggest that other domains in Las17 outside of VCA contribute to its robust NPF activity. One recent study identified sequences in the proline-rich (PR) region of Las17, adjacent to the VCA, that bind actin and contribute to NPF activity (Urbanek et al. 2013). However, the NPF activity of this PR-VCA fragment was still much weaker than that of full-length Las17, suggesting that there is still much to learn about Las17 activity. Importantly, this study also reported that some Las17-stimulated actin nucleation was Arp2/3 independent; such activity could contribute to filament initiation at the patch (see below). Purified Las17 is also constitutively active rather than autoinhibited like mammalian N-WASP (Rodal et al. 2003). So, in this regard Las17 is more similar to mammalian SCAR/WAVE proteins, other members of the larger WASP/SCAR/WAVE family (Welch and Mullins 2002). WAVE proteins are inhibited in trans by a large multisubunit WAVE-regulatory complex (WRC) (Eden et al. 2002; Gautreau et al. 2004; Chen et al. 2010). Although homologs of WRC genes are not found in yeast, a distinct set of Las17-binding partners appears to mediate trans-inhibition (see below). Therefore, while the domain layout of Las17 resembles WASP, its regulation more closely resembles SCAR/WAVE.
Myosin I (Myo3 and Myo5):
Type I myosins are considered unconventional within the myosin superfamily because they are single headed, nonprocessive, and lipid associated. There are two yeast homologs of myosin I, called Myo3 and Myo5. Each has a motor domain, lipid and F-actin binding domains, an SH3 domain, and an Arp2/3-interacting acidic (A) domain (Figure 6C). Myosin I NPF activities require both the A domains and interactions of their SH3 domains with Vrp1/WIP, which in turn binds G-actin (Sirotkin et al. 2005; Sun et al. 2006). Thus, the combination of myosin I and Vrp1 provides a functional equivalent to VCA and delivers actin monomers to Arp2/3 complex. Single deletions of MYO3 or MYO5 cause relatively minor defects in vivo, but double mutants show pronounced defects in endocytosis and actin organization (Geli and Riezman 1996). Both the motor and NPF activities of myosin I are required for its endocytic functions (Sun et al. 2006).
In addition to its roles as a scaffold protein in organizing other components of the endocytic machinery (see Cargo Selection and Coat Formation; Figure 2), Pan1 contains an Arp2/3-interacting A domain and an unconventional WH2 (V) domain that binds F-actin rather than G-actin (Figure 6). Pan1 has weak NPF effects on Arp2/3 complex compared to those of Las17 or myosin I and Vrp1. The NPF activities of Pan1 depend on its WH2 and A domains (Duncan et al. 2001; Toshima et al. 2007). Pan1 is categorized as a class II NPF (Welch and Mullins 2002), but its functions in regulating Arp2/3 complex are still not well understood.
Abp1 (actin binding protein 1) was first isolated from yeast-cell extracts on an F-actin affinity column (Drubin et al. 1988) and later found to be conserved in mammals (Ensenat et al. 1999; Larbolette et al. 1999; Kessels et al. 2000). It has an N-terminal actin depolymerization factor (ADF)/cofilin-homology (ADFH) domain through which it binds F-actin but not G-actin and two acidic (A) domains that bind Arp2/3 complex (Kessels et al. 2000; Goode et al. 2001) (Figure 6C). The C-terminal half of Abp1 contains a proline-rich region that binds the SH3 domain of Rvs167 (Friesen et al. 2006) and an SH3 domain that binds Srv2, Scp1, and the kinases Ark1 and Prk1 (Freeman et al. 1996; Lila and Drubin 1997; Fazi et al. 2002; Haynes et al. 2007); the functions of these proteins are described below. Like Pan1, the NPF effects of Abp1 are weak compared to those of class I NPFs (Sun et al. 2006). Further, because Abp1 binds to F-actin, it is categorized as a class II NPF. Although abp1∆ mutants have no discernable defects in actin organization and only subtle defects in endocytosis (Lila and Drubin 1997; Kaksonen et al. 2005), double mutants with sla1, sla2, or sac6 are lethal (Holtzman et al. 1993), demonstrating that the functions of Abp1 are important, although genetically redundant. Interestingly, all three shared functions of Abp1 depend on its SH3 domain, but only two of these (shared with Sla2 and Sac6) depend on its ability to bind F-actin (Quintero-Monzon et al. 2005). Recently, Abp1 was found to be essential for an alternative endocytic pathway in yeast, perhaps related to its synthetic lethality with factors central to the conventional CME pathway (Aghamohammadzadeh et al. 2014). While purified Abp1 alone shows no effects on actin-filament dynamics or organization in vitro (Goode et al. 2001; Quintero-Monzon et al. 2005), when bound to its ligand Aim3 it can cap actin-filament barbed ends (Michelot et al. 2013). Through this interaction, and its interactions with the Arp2/3 complex and the other binding partners mentioned above, Abp1 may influence actin-filament assembly and turnover in a number of ways.
Regulation of actin nucleation in endocytosis:
How are the activities of Arp2/3 complex’s four NPFs utilized and controlled during endocytosis? As mentioned above, the precise spatial recruitment and/or localized activation of NPFs at endocytic sites is likely to be critical in enabling the polymerizing actin network to provide the appropriate directional force to drive plasma membrane invagination and tubulation. Recent EM studies have begun to define the locations of the NPFs to give us clues to how this works (Idrissi et al. 2008; Idrissi et al. 2012). These studies suggest that two different pools of NPFs, one at the plasma membrane (Las17, Myo3, and Myo5) and the other at the tip of the invagination (Pan1 and Abp1), might control two different phases of endocytosis, invagination and scission, respectively.
In addition to the proper spatial recruitment of NPFs at endocytic sites, it is critical that their activities be tightly controlled in a temporal manner. Las17 and Pan1 are the first NPFs to appear at patches, arriving at least 20 sec before the onset of actin assembly, and subsequently myosin I and Abp1 arrive shortly after the first appearance of F-actin. These observations beg the question, How are early arriving NPFs, Las17 and Pan1, kept inactive at endocytic sites until the required onset of actin assembly?
Control of early NPF localization and activity:
Among the many binding partners of Pan1, so far only Sla2 has been shown to regulate Pan1 NPF activity, inhibiting its effects in vitro (Toshima et al. 2007). This may be critical for blocking actin assembly until an endocytic patch has sufficiently matured, although much work is still needed to fully understand Pan1 regulation in vivo. In contrast, our understanding of Las17 regulation has expanded significantly in recent years. Five direct inhibitors of Las17 activity have been identified: Sla1, Bbc1, Syp1, Lsb1, and Lsb2 (Rodal et al. 2003; Boettner et al. 2009; Spiess et al. 2013). Syp1 arrives at patches well before Las17, and Sla1 arrives in approximately the same time frame as Las17, making these binding partners prime candidates for constraining Las17 activity (Kaksonen et al. 2003; Stimpson et al. 2009). In contrast, Bbc1 arrives later, suggesting that it may have a different role than that of a Las17 inhibitor, possibly in spatially restricting Las17 activity during vesicle internalization. Finally, the function and localization of Lsb1 and Lsb2 are unclear, but overexpressed Lsb1 redistributes Las17 into cytoplasmic puncta, and together with the Ub-ligase Rsp5 and Sla1, Lsb1 is implicated in the regulation of Las17 protein levels (Kaminska et al. 2011; Spiess et al. 2013). As mentioned earlier, Syp1 binds and inhibits Las17, which is achieved through its central region (Boettner et al. 2009). Sla1 contains multiple SH3 domains through which it binds to proline-rich motifs in Las17 to inhibit its NPF activity (Rodal et al. 2003; Feliciano and Di Pietro 2012). Syp1 and Sla1 also directly associate to form a stable complex in cells and together more effectively block Las17 activity in vitro (Feliciano and Di Pietro 2012) (Figure 6A). It should be noted that in biochemical assays testing inhibition of Las17, high ratios of the inhibitors to Las17 (likely higher than those found at endocytic sites in vivo) are required to block activity. However, this does not rule out the possibility that clustering at the membrane, lipid/scaffold protein binding, and/or post-translational modifications in vivo drive these interactions to improve the efficiency of Las17 inhibition.
How is inhibition of Las17 by Syp1 and Sla1, and possibly Lsb1 and Lsb2, subsequently released to allow actin assembly? The first clue is the observation that Syp1 disappears from patches just before the onset of actin assembly (Boettner et al. 2009; Stimpson et al. 2009). The mechanism for promoting Syp1 departure from patches is unknown, but one possibility is direct phosphorylation by Ark1 and Prk1 kinases (Huang et al. 2003; Albuquerque et al. 2008). The second clue is the arrival Bzz1 a few seconds before the onset of actin assembly (Sun et al. 2006). Bzz1 binds Las17 via its SH3 domains and is predicted to compete with Sla1 (Soulard et al. 2005; Tonikian et al. 2009). Further, S. pombe Bzz1 stimulates WASP–Arp2/3-mediated actin assembly in vitro, and mutants have reduced number of actin patches and impaired patch dynamics (Arasada and Pollard 2011). Bzz1 dimerizes using its F-BAR domain and thus could elevate Las17 NPF activity by organizing two Las17 molecules (Figure 6C). It is tempting to speculate that the combination of Syp1 departing and Sla1 facing new competition from Bzz1 is sufficient to trigger Las17 release at the base of the invagination, allowing it to recruit Arp2/3 complex and initiate actin assembly. Consistent with this model, mutations in Bzz1 and myosin I are synthetic lethal (Soulard et al. 2002), as are mutations in Las17 and myosin I.
Recent studies combining light and electron microscopy have shown that Bzz1 is recruited to endocytic sites when membrane curvature is low, remains at the base of the invagination, and helps shape the tubule in coordination with the BAR domain proteins Rvs161 and Rvs167 (Kishimoto et al. 2011). Like F-BAR proteins, Bzz1 may self-assemble to form a rigid collar at the base of the invagination, providing resistance to actin-assembly forces, while Rvs proteins coat the tubule to help shape the vesicle. Bzz1 may also locally activate Las17 at the base of the invagination, directly coordinating membrane remodeling and actin assembly. Similar roles have been suggested for Sla2 and Rvs167 in coordinating membrane and actin remodeling; Sla2 links membranes to F-actin, Sla1, Vrp1, and Ysc84, while Rvs167 links membranes to Las17, Ysc84, Abp1, and Aim3.
The search for Eve: the first mother filament:
Even after Las17 is activated, there is another requirement for initiation of an actin network by Arp2/3 complex, which is a first mother filament, or alternatively a mechanism for bypassing the requirement of a first filament. Two different models have been put forth to explain how this might be achieved. First, it has been suggested that short filament fragments produced by severing events at nearby patches or cables may diffuse to new sites of endocytosis and be captured to prime network assembly. This model is supported by one recent study analyzing the effects of cofilin mutants on patch formation (Chen and Pollard 2013). Second, it has been postulated that there is specialized machinery at patches that either (a) induces conformational changes in Arp2/3 complex similar to mother filament binding or (b) recruits and organizes actin monomers into a short F-actin seed that serves as the primer (Figure 6B). In this regard, as noted above, full-length Las17 has NPF activity at least 2 orders of magnitude stronger than just its VCA region (Rodal et al. 2003), suggesting that other domains in Las17 might contribute to such a function. Further, the Las17-binding partner Ysc84/Lsb4 binds F-actin and together with Las17 (no Arp2/3) can stimulate actin polymerization (Robertson et al. 2009). Further the S. pombe WISH/DIP/SPIN90 ortholog Dip1 (Ldb17 in S. cerevisiae) binds Arp2/3 complex and stimulates unbranched nucleation, suggesting that in vivo it may trigger formation by Arp2/3 complex of the first mother filament to initiate actin network assembly, a model that is strongly supported by analysis of its mutant phenotype (Basu and Chang 2011). Interestingly, Ldb17 has interactions with Sla1, Bzz1, and Ysc84/Lsb4, which further suggests that a number of these factors may contribute to a multicomponent regulatory mechanism controlling both the initial restraint and subsequent activation of actin network formation.
Orchestration of multiple NPFs:
The next NPFs to arrive at sites of endocytosis, after Pan1 and Las17, are Vrp1 and myosin I, in that order. Live-imaging analysis has shown that Las17 recruits Vrp1, which together with F-actin in turn recruits myosin I to patches (Soulard et al. 2002; Jonsdottir and Li 2004; Sirotkin et al. 2005; Sun et al. 2006). Current models suggest that myosin I remains anchored to the plasma membrane, while its nucleation activity and motor activity cooperate to produce directional force in facilitating membrane remodeling. Specifically, its barbed end-directed motor activity at the base of the invagination could provide an inward force on the entire actin network to drive tubulation. To facilitate this step, there may be patch components that help cluster and anchor myosin I at the base; one candidate for this role is Pan1, which binds myosin I through its C-terminal proline-rich domain (PRD) and increases the combined NPF activity of myosin I and Vrp1 (Barker et al. 2007). How the motor activity of myosin I is coordinated with its NPF effects and if/when myosin I functions in a complex with Las17 remain open questions. Las17 and myosin I activities also may not be restricted to the base of the patch, as originally thought, since more recent studies (using new tags that are possibly more functional) suggest that Las17 and myosin I might be positioned not only at the base of the patch but also at the tip of the invagination (Idrissi et al. 2008; Idrissi et al. 2012). What role these NPFs play at the tip of the invagination is unclear, but may involve molding the tubular invaginations into more bulbous vesicles (Kukulski et al. 2012a).
Last to arrive at patches is Abp1, another poorly understood NPF. abp1 mutants show a pattern of genetic interactions with other NPFs and arp2 alleles that suggests an antagonistic relationship with class I NPFs (D’agostino and Goode 2005). Consistent with this model, purified Abp1 competes directly with Las17 VCA region for binding Arp2/3 complex and attenuates VCA stimulatory effects. However, this may not be the full extent of Abp1’s Arp2/3-regulating functions. Much like myosin I, which had little effect on Arp2/3 activity until it was combined with its ligand Vrp1 (Sun et al. 2006), Abp1 may have important effects on Arp2/3 complex that require the presence of other NPFs and/or Abp1 binding partners.
Actin-filament capping, cross-linking, and stabilization
Once nucleated, daughter filaments in the actin network must be capped rapidly at their barbed ends to (a) restrict filament length and produce a network consisting of densely branched, short interconnected filaments and (b) maintain the actin monomer pool in cells to support rapid growth at new sites of nucleation (Loisel et al. 1999; Blanchoin et al. 2000) (Figure 4). Thus, barbed end capping plays an instrumental role in enabling networks to generate directional force applied to membrane protrusion in motility and membrane invagination in endocytosis. In addition to being capped, newly formed filaments are decorated along their sides by a number of actin-associated proteins that stabilize, crosslink, and/or bundle F-actin. These proteins may be critical for holding the network together, and for transferring mechanical forces from actin polymerization and myosin activity across the network to deform the membrane (Carlsson and Bayly 2014).
For years, it was presumed that capping of Arp2/3 complex-generated branched networks was mediated solely by heterodimeric “capping protein,” Cap1/Cap2 (CP), which has a well-defined structure and interactions with the filament end (reviewed in Cooper and Sept 2008). However, this left the nagging question—Why do mutations in CP cause only minor defects in actin patch organization or dynamics (Amatruda et al. 1990; Kim et al. 2006)? New insights into this question have emerged from genetic and biochemical studies demonstrating that the combination of the disassembly factors Aip1 and Cof1 (see below) blocks barbed end growth (Okada et al. 2002) and that mutations in CP and Aip1 are synthetic lethal (Balcer et al. 2003). Further, as mentioned above, a recent study showed that the combination of Abp1 and its ligand Aim3 caps barbed ends and that mutations in either Abp1 or Aim3 display synthetic genetic interactions with CP and Aip1 (Michelot et al. 2013). In addition, this study showed that CP and Aip1 localize to distinct regions of actin networks (newer and older, respectively, as visualized in the elongated actin structures of sla2∆ mutants) and that when CP is deleted Aip1 spreads into CP territory. Further, a complementary study in S. pombe shows that CP spreads into Aip1 territory when AIP1 is deleted (Berro and Pollard 2014). Thus, actin patches have at least three different barbed end-capping mechanisms (CP, Aip1–Cof1, and Abp1–Aim3), although there is much more to learn about their specific roles and mechanisms.
Intriguingly, the phosphoinositide PtdIns(4,5)P2 (which is critical for endocytosis; see Vesicle Scission and Uncoating) can weaken interactions between CP and barbed ends in other systems (Schafer et al. 1996). Since barbed ends are thought to face the membrane at the base of the endocytic invagination, protection against capping at these locations by PtdIns(4,5)P2 could allow polymerization toward the membrane and thus direct expansion of the actin network in the right orientation.
Different F-actin crosslinking proteins can organize filaments into parallel and antiparallel bundles or orthogonal arrays, giving rise to networks of filaments with distinct geometries and mechanical properties (Bartles 2000; Pollard and Cooper 2009). Yeast cells have at least five different actin crosslinking proteins that potentially regulate the spatial organization of filaments at sites of endocytosis, Sac6/fimbrin, Scp1/calponin, Abp140, Crn1, and Tef1/Tef2, each discussed in detail below. As mentioned above, it is not yet clear where crosslinking of filaments occurs within the presumably densely branched actin patch networks. One possibility is that debranching effects rapidly remodel specific regions of a patch (see below), and then crosslinking proteins reorganize filaments into bundles or orthogonal arrays. Another possibility is that crosslinks are added to reinforce connections among already branched filaments, providing additional mechanical strength to the network (as illustrated in Figure 4). Since at least five crosslinking proteins potentially act on distinct regions of an actin patch, they may function during specific temporal phases as the patch matures, elongates, and moves away from the plasma membrane. Further, their individual activities may help generate distinct architectural zones within the network during any given phase.
Sac6 was first isolated from yeast-cell extracts on F-actin affinity columns (Drubin et al. 1988) and in parallel in genetic screens as a suppressor of an act1 allele (Adams et al. 1989). It is the yeast homolog of the widely expressed protein fimbrin, which bundles filaments into parallel arrays in vitro (Bretscher 1981; Glenney et al. 1981). The crystal structures of Arabidopsis thaliana and S. Pombe fimbrin show that it consists of four tandem calponin homology (CH) domains organized into two separate actin-binding units (Klein et al. 2004). The 3D reconstructions from electron micrographs of fimbrin-decorated F-actin bundles suggest a regular pattern of crosslinks with a specific geometry (Hanein et al. 1998; Volkmann et al. 2001). However, there is currently only limited evidence to support a role for yeast Sac6 functioning as a bundler in vivo. One study found that a truncated S. pombe Sac6/fimbrin defective in bundling F-actin in vitro failed to rescue fimbrin function at actin patches in vivo (Skau et al. 2011), suggesting that bundling might be essential for fimbrin in vivo function. However, this truncation mutant reduced actin-binding affinity, leaving open alternative interpretations. There are at least two other biochemical activities of Sac6/fimbrin: (i) F-actin stabilization and (ii) competitive inhibition of tropomyosin decoration of F-actin at patches (Goodman et al. 2003; Skau and Kovar 2010). Thus, more work is needed to determine conclusively whether Sac6/fimbrin, or any of the other in vitro bundling proteins (below), actually crosslink filaments in vivo at actin patches.
Scp1 is the yeast homolog of the mammalian actin-bundling protein calponin (22 kDa), also referred to as SM22 or transgelin. Vertebrate genomes contain three calponin genes, which have important functions in both muscle and nonmuscle cells (Rozenblum and Gimona 2008; Wu and Jin 2008). Yeast Scp1 is composed of an N-terminal CH domain, a short proline-rich segment, and a single C-terminal calponin-like repeat (CLR). Although CH domains bind actin in many proteins (e.g., Sac6/fimbrin and α-actinin), the CH domain of calponin does not (Goodman et al. 2003) and instead may bind microtubules and/or other ligands (Rozenblum and Gimona 2008). The C-terminal half of Scp1 mediates actin bundling and harbors two distinct actin-binding sequences, one in the proline-rich region and one in the CLR (Goodman et al. 2003; Winder et al. 2003; Gheorghe et al. 2008). Although scp1∆ mutants alone do not cause obvious defects in actin organization or endocytosis, scp1∆ sac6∆ double mutants exhibit gross abnormalities in cell size, actin organization, and endocytosis, far more severe than defects in sac6∆ mutants (Goodman et al. 2003). Further, overexpression of Scp1 partially rescues the single-mutant defects of sac6∆. Together, these observations suggest that Scp1 and Sac6 perform related, overlapping functions in vivo, which is consistent with biochemical observations showing that each protein bundles and stabilizes F-actin. Sac6 is also upregulated in scp1∆ mutants, which may explain why the scp1∆ phenotype is not stronger (Gheorghe et al. 2008). Live-cell imaging analysis has revealed that Scp1 helps promote inward movement of patches from the plasma membrane (Gheorghe et al. 2008). These observations raise the intriguing possibility that Scp1 and Sac6 play a role in further crosslinking Arp2/3-generated branched networks and stiffened them to facilitate myosin I-dependent force transmission during invagination and scission.
Abp140 was isolated from yeast cell extracts in an assay for novel F-actin bundling factors, and localizes to both actin patches and cables (Asakura et al. 1998). Although its molecular weight is 70 kDa, it was initially thought to have a molecular weight of 140 kDa (likely due to disulfide bond formation), giving rise to its name. To date, Abp140–GFP has provided the only effective marker for live-cell imaging of actin cables (Yang and Pon 2002). In fact, the popular actin marker Lifeact, used to label actin in live-cell imaging studies in a wide range of organisms, is a small peptide derived from Abp140 (Riedl et al. 2008). Because abp140∆ causes no obvious change in actin organization or dynamics, its role in the actin cytoskeleton may be redundant with other actin-binding proteins, but this has not been explored genetically. More recent studies have revealed that in addition to binding actin, Abp140 is a methyltransferase that modifies the tRNA anticodon loop (D’silva et al. 2011; Noma et al. 2011). Further, ABP140 mRNA is tethered to nascent Abp140 protein, which binds actin and moves rapidly with actin cable retrograde flow to reach the back of the mother cell, accumulating at this pole (Kilchert and Spang 2011). These observations provide intriguing new links between actin and protein translational control, adding to the wealth of previous observations connecting actin and translation (Hesketh 1994).
Crn1 is the yeast homolog of coronin and binds with high affinity to F-actin and homo-oligomerizes via its C-terminal coiled-coil domain to crosslink actin filaments in vitro (Goode et al. 1999). The actin-binding surfaces on Crn1’s N-terminal β-propeller domain have been mapped to evolutionarily conserved residues and are required for its shared genetic functions with Cof1 (Gandhi et al. 2010a). Genetic interactions for crn1∆, as well as the late arrival of Crn1 to actin patches (Lin et al. 2010), suggest that its primary role in actin regulation may be to promote turnover (see below); thus, more work is needed to determine whether the actin crosslinking activity is relevant to Crn1’s in vivo functions.
Tef1 and Tef2 are the yeast homologs of protein translation elongation factor eEF1A (formerly known as eEF1α), which is one of the most abundant proteins in eukaryotic cells and delivers aminoacyl-tRNA to translating ribosomes. Tef1 and Tef2 are identical in sequence and genetically redundant. In addition to its canonical role, Tef1/Tef2/eEF1A has an important role in F-actin bundling (Murray et al. 1996; Munshi et al. 2001). This function depends on a specific actin-binding surface, and mutations at this site abolish actin interactions without affecting protein translation (Gross and Kinzy 2005). Using these alleles, it was shown that Tef1-actin interactions are important for the proper formation of actin cables, which is also consistent with reported interactions between Tef1 and the formin Bni1 (Umikawa et al. 1998). Because of its abundance in the cytosol, it has been difficult to determine whether a subpopulation of Tef1 localizes to cables and/or patches, and there have been no investigations into the potential role of Tef1 in endocytosis.
Actin-filament disassembly and turnover
About 3–7 sec after actin assembly is initiated at the cortical patches, components of the disassembly and turnover machinery begin arriving (Okreglak and Drubin 2007; Lin et al. 2010). During the next 10–15 sec of the actin patch lifetime, there is continued actin assembly, likely concurrent with disassembly, which predicts a highly dynamic actin network. Maintaining a high rate of turnover may endow the actin network with plasticity, which may be especially important in processes such as endocytosis and lamellipodial protrusion, where the shape of the actin network changes dramatically with membrane invaginations or protrusions. In endocytosis, the shape and architecture of the actin network is likely to change drastically over the course of its 15-sec lifespan (Figure 1, G and H). Based on the measured rate of actin turnover in extended cortical actin patches (0.05 µm sec−1, or about 70 subunits sec−1) (Kaksonen et al. 2003), it is possible that the entire F-actin content of a patch turns over three to four times during the patch lifetime. This begs the question—What are the mechanisms that enable such rapid actin turnover, and how are they balanced with new actin assembly, to maintain a network of a particular size while allowing it to rapidly evolve in shape and architecture?
From knowledge gained over the past 17 years, an answer to the question above is that actin-filament turnover is controlled by two major factors: (i) the changing nucleotide state (age) of filaments and (ii) a complex set of machinery that responds to these changes and works collectively to efficiently dismantle older regions of filaments without affecting newer regions. Below, we first discuss the nucleotide state of F-actin and the components of the actin turnover machinery (Figure 7) and then provide an overview of their combined mechanisms in F-actin severing and depolymerization and G-actin recycling (Figure 4).
The nucleotide state of actin:
Critical to the control of F-actin disassembly is the age of the filament, specified by its nucleotide state (Pollard et al. 2000; Dominguez and Holmes 2011). When each ATP-actin monomer is added to the barbed end of a filament, its noncovalently bound ATP undergoes very rapid hydrolysis to generate ADP and inorganic phosphate (Pi). Hydrolysis occurs at a rate of 0.3 sec−1 for rabbit muscle actin, and possibly much faster for yeast actin (Bryan and Rubenstein 2005). The products remain bound to actin, as ADP + Pi-actin, and then in a subsequent and slower step, Pi is released from the filament to yield ADP-actin. The rate of Pi release for muscle actin is 0.0022 sec−1, but has not been measured for yeast actin. Recent results show that each subunit in a filament formed by muscle actin releases Pi independently, without influence from its neighboring subunits (Jegou et al. 2011). Thus, each filament is thought to be a unique mosaic of subunits in different nucleotide states, with the growing end of the filament enriched in ATP and ADP + Pi-actin, and the older end of the filament enriched in ADP-actin. The relative amounts and distribution of each form of actin will depend on at least four factors: (a) the filament length, (b) the filament growth rate, (c) the species-specific rate constants for hydrolysis and Pi release, and (d) the influence of actin-binding proteins on these rates. Importantly, F-actin in each nucleotide state has a distinct conformation, which endows it with different dynamic properties and recruits a different set of actin-associated proteins (Dominguez and Holmes 2011). One unanswered question is what portion of a yeast endocytic actin network exists in each nucleotide state, but this is difficult to answer in the absence of more information on points c and d above. Clearly, this is an area ripe for future investigation.
The actin turnover machinery:
Below, we discuss individually each of the seven known, and highly conserved, components of the actin turnover machinery (cofilin, GMF, Aip1, coronin, Srv2/CAP, profilin, and twinfilin), focusing on their protein structures, biochemical activities, and mutant phenotypes in yeast. We then discuss how the activities of these components are orchestrated to facilitate each step in actin turnover (Figure 4 and Figure 7).
Cofilin, also called ADF, is one of the more abundant actin-associated proteins in yeast and mammalian cells. It is an 18-kDa globular protein that binds with higher affinity to ADP-actin monomers and filaments and lower affinity to ATP-actin (reviewed in Ono 2007). Cofilin binds cooperatively to F-actin (Blanchoin and Pollard 1999), leading to an uneven pattern of decoration that can be visualized by two-color TIRF microscopy (Suarez et al. 2011; Chaudhry et al. 2013). Decoration by cofilin “twists” filaments (McGough and Chiu 1999) and causes major conformational rearrangements (Galkin et al. 2011). Severing occurs at the boundaries between cofilin-decorated and undecorated regions on filaments due to mechanical discontinuities (Suarez et al. 2011). However, it remains to be seen how the severing mechanism is affected when cofilin functions in concert with other actin disassembly factors (below). The COF1 gene is essential (Moon et al. 1993), but partial loss-of-function cof1 alleles generated by scanning mutagenesis (Lappalainen et al. 1997) show defects in patch dynamics just before scission, pointing to an important role for cofilin-mediated actin turnover and remodeling in driving invagination (Lappalainen and Drubin 1997; Okreglak and Drubin 2007, 2010; Lin et al. 2010). Additional defects are observed at later stages of endocytosis, including a delayed uncoating of actin from internalized vesicles.
Gmf1, the yeast homolog of mammalian GMF (glia maturation factor), is an 18-kDa protein with structural similarity to cofilin (Goroncy et al. 2009). GMF is one of four proteins in the ADFH domain superfamily conserved from yeast to humans: cofilin, Abp1, twinfilin, and GMF (Poukkula et al. 2011). Gmf1 has little if any binding affinity for F-actin, but binds strongly to Arp2/3 complex (Figure 7) and genetically interacts with cofilin suggesting a role in turnover (Ikeda et al. 2006; Gandhi et al. 2010b; Nakano et al. 2010). Gmf1 catalyzes debranching of filaments in vitro without severing filaments at nonbranch locations (Gandhi et al. 2010b). The debranching activity of yeast Gmf1 is conserved in Drosophila and mouse GMFs (Poukkula et al. 2014) and depends on separate binding interactions with Arp2 and the first actin subunit in the daughter filament, suggesting a modified cofilin-like severing mechanism for debranching (Luan and Nolen 2013; Ydenberg et al. 2013). GMF preferably binds to ADP-rather than ATP-Arp2/3 complex (Boczkowska et al. 2013), suggesting that it targets “aged” branches. GMF also inhibits Arp2/3-mediated actin nucleation by competing with VCA for binding Arp2/3 (Gandhi et al. 2010b; Nakano et al. 2010; Ydenberg et al. 2013). The in vivo relevance of this latter activity is unclear, but could help inhibit branched nucleation by Arp2/3 complex in the absence of NPFs or suppress branched nucleation at the older (ADP-rich) regions of networks to allow filament crosslinkers to act.
Aip1 (actin-interacting protein 1) was first identified in a two-hybrid screen for novel interaction partners of yeast actin (Amberg et al. 1995) and subsequently as a binding partner of cofilin (Rodal et al. 1999). A deletion of AIP1 causes only mild phenotypes, but is synthetic lethal with specific cof1 alleles. Purified Aip1 alone has no effect on actin-filament dynamics, but in combination with cofilin promotes the net disassembly of F-actin into monomers and short oligomers (Okada et al. 1999; Rodal et al. 1999; Okreglak and Drubin 2010). The mechanism underlying these effects is thought to involve enhanced severing and blocking barbed end growth at severed ends (Okada et al. 2002; Balcer et al. 2003; Ono et al. 2004; Brieher et al. 2006; Tsuji et al. 2009; Michelot et al. 2013). Aip1 consists of two β-propellers in one polypeptide chain and its structure has an open clamshell appearance (Voegtli et al. 2003; Mohri et al. 2004) (Figure 7). The actin- and cofilin-binding surfaces on Aip1 have been defined by mutagenesis (Mohri et al. 2004; Clark et al. 2006; Okada et al. 2006) and are required for Aip1 functions in vivo in promoting rapid turnover of actin patches and cables (Okada et al. 2006; Okreglak and Drubin 2010). Purified Aip1 binds weakly to F-actin unless it is decorated with cofilin (Okada et al. 1999; Rodal et al. 1999; Mohri et al. 2004; Okada et al. 2006), and Aip1–GFP recruitment to actin patches in vivo depends on cofilin (Lin et al. 2010). Recent biochemical, structural, and cell biological evidence has suggested that Aip1 may enhance severing by competing with Cof1 for its second binding site on F-actin, leading to filament destabilization. Indeed, in support of this model, mutants in cof1 that disrupt its second F-actin binding site were found to be hyperactive for filament severing (Aggeli et al. 2014).
Crn1, the yeast homolog of coronin, binds with high affinity to ATP-F-actin and much lower affinity to ADP-F-actin (Goode et al. 1999; Cai et al. 2007; Gandhi et al. 2009). It contains an N-terminal β-propeller domain (Appleton et al. 2006), a central unique sequence (highly distinct across distant species) and a C-terminal coiled-coil domain, through which most coronins trimerize (De Hostos 1999) (Figure 7). Crn1 has two separate actin binding sites, one in its β-propeller domain (Gandhi et al. 2010a) and another in its coiled-coil domain (Gandhi et al. 2009). A crn1∆ mutation alone causes no obvious defects in actin organization or endocytosis (Heil-Chapdelaine et al. 1998; Goode et al. 1999), but exhibits synthetic genetic interactions with cof1-22 and act1-159 alleles, suggesting a role in actin turnover (Goode et al. 1999; Gandhi et al. 2009). Consistent with this view, purified Crn1 promotes severing of ADP–actin filaments by cofilin in vitro (Gandhi et al. 2009), and in mammalian systems, coronin, cofilin, and Aip1 have been coisolated from cell extracts as a three-component system that promotes the rapid disassembly of ADP-filaments (Brieher et al. 2006; Kueh et al. 2008). The mechanisms underlying these effects for either yeast or mammalian coronins are not yet defined and require further investigation. Adding to the potential complexity of coronin’s roles in actin turnover are the interactions of its unique sequence and C-terminal coiled-coil domain with Arp2/3 complex (Humphries et al. 2002; Liu et al. 2011).
Srv2, also called CAP (cyclase-associated protein), was discovered in yeast as a suppressor of a hyperactivated ras allele and as a binding partner of membrane-associated adenylate cyclase (Fedor-Chaiken et al. 1990; Field et al. 1990). Srv2/CAP is a bifunctional protein. Its N-terminal half (N-Srv2) hexamerizes into shuriken-like structures that associate with cofilin-decorated F-actin and improve the efficiency of severing (Normoyle and Brieher 2012; Chaudhry et al. 2013) (Figure 4 and Figure 7). Point mutations in N-Srv2 that abolish the enhanced severing activity in vitro also cause striking defects in actin organization in vivo. In contrast to these functions, the C-terminal half of the protein (C-Srv2) binds to G-actin, and has high affinity for ADP-G-actin (Kd 18 nM) but 100-fold lower affinity for ATP-G-actin (Mattila et al. 2004). This enables C-Srv2 to potently catalyze recycling of cofilin from ADP-actin monomers and nucleotide exchange on actin (ATP for ADP). These activities depend on the coordinated effects of two separate actin-binding sites in C-Srv2, and mutations that disrupt these G-actin interactions in vitro also cause defects in actin organization in vivo (Mattila et al. 2004; Chaudhry et al. 2010). C-Srv2 also has a binding site for profilin, which may enable Srv2 and profilin to work together in monomer recycling (Bertling et al. 2007). One recent study showed that the functions of the N and C termini of Srv2 are autonomous, such that the two physically separated halves can function in trans both in vitro and in vivo (Chaudhry et al. 2014). However, it was suggested that linkage of the two halves in a single polypeptide may be important for coordinating the two activities in vivo. Finally, the structure, activities, and mechanism of this protein were recently shown to be strikingly similar between the yeast and mammalian homologs of Srv2/CAP (Normoyle and Brieher 2012; Jansen et al. 2014), demonstrating that its roles in actin regulation have been conserved in evolution.
Profilin is a conserved 15-kDa protein that binds preferentially to ATP-actin monomers and also catalyzes nucleotide exchange on actin (ATP for ADP) (Blanchoin and Pollard 1998; Eads et al. 1998; Balcer et al. 2003). Profilin suppresses spontaneous nucleation of actin and blocks monomer addition to pointed ends while freely allowing addition to barbed ends. As a result, profilin helps to restrict new actin growth in vivo to barbed ends generated by nucleation mechanisms. Profilin also binds to specific polyproline-rich tracts in other actin regulatory proteins, such as formins and Srv2 (Chang 1999; Bertling et al. 2007) (Figure 4 and Figure 7). The best understood interaction of profilin is with formins, where it plays a critical role in delivering actin monomers to accelerate barbed end elongation. (Kovar et al. 2006; Paul and Pollard 2008, 2009). The actin- and polyproline-binding surfaces on profilin are well defined (Schutt et al. 1993; Eads et al. 1998), and mutations at these sites in yeast profilin (PFY1) cause severe depolarization and misorganization of actin structures (Haarer et al. 1993; Wolven et al. 2000; Lu and Pollard 2001), including a striking loss of actin cables, consistent with its role in formin-mediated actin assembly (Sagot et al. 2002; Kovar et al. 2006). Further, at least one of the actin-binding-deficient pfy1 mutants displays defects in fluid-phase endocytosis (Wolven et al. 2000). Since there is no evidence supporting a role for profilin in promoting Arp2/3-dependent assembly of branched actin-filament networks at sites of endocytosis, the potential connections between PFY1 and endocytosis have remained elusive. However, one possibility is that the loss of actin cables resulting from mutations in profilin indirectly compromises endocytosis. Finally, it should be noted that profilin binds PtdIns(4,5)P2 and that this interaction competes with actin binding, raising the possibility that local changes in membrane composition influence profilin’s effects on the actin cytoskeleton (Lassing and Lindberg 1985; Ostrander et al. 1995).
Twf1 (or twinfilin) was identified in yeast both on F-actin affinity columns and in database searches for proteins with homology to Cof1 (Goode et al. 1998). Twinfilin is composed of two ADFH domains connected by a short linker sequence and has a short C-terminal tail region (Figure 7). Studies on yeast and mouse twinfilin have shown that the N-terminal ADFH domain binds G-actin, whereas the C-terminal domain binds G- and F-actin (Paavilainen et al. 2002; Hellman et al. 2006; Paavilainen et al. 2007; Paavilainen et al. 2008). In addition, the C-terminal tail of yeast twinfilin binds to CP, and this interaction is critical for Twf1 localization to actin patches and Twf1’s shared genetic functions with Cof1 (Palmgren et al. 2001; Falck et al. 2004). However, Twf1 has no detectable effect on CP activity in vitro, nor does CP have any detectable effects on Twf1 interactions with G-actin, which has left the functional role of the Twf1–CP interaction unclear. Yeast and mammalian twinfilin proteins bind with high affinity to ADP-G-actin and suppress monomer addition to filament ends (Goode et al. 1998; Vartiainen et al. 2000), which led to early models suggesting that twinfilin serves primarily as an actin monomer sequestering protein similar to thymosin-β4 (Palmgren et al. 2002; Poukkula et al. 2011). However, other lines of evidence have pointed to twinfilin-promoting actin-filament disassembly or turnover, including: (i) twinfilin genetically interacts with cofilin in yeast and flies (Goode et al. 1998; Wahlstrom et al. 2001); (ii) genetic perturbation of twinfilin in yeast cells slows actin turnover, and in mammalian cells slows cell motility (Moseley et al. 2006; Meacham et al. 2009; Wang et al. 2010); (iii) twinfilin localizes to sites of dynamic actin turnover in these systems: patches in yeast, and the leading edge or endosomal actin tails in mammalian cells (Palmgren et al. 2001; Helfer et al. 2006); (iv) the C-terminal ADFH domain in twinfilin binds F-actin in addition to G-actin and can interact with the barbed end of the filament (Paavilainen et al. 2007); and (v) Twf1 associates with the actin turnover-promoting protein Srv2 in vivo (Tarassov et al. 2008; Breitkreutz et al. 2010). Thus, we have included twinfilin here, with the other actin turnover machinery; however, the specific role and mechanism of twinfilin in promoting turnover remains a mystery.
Coordinated mechanisms of actin-filament disassembly and turnover:
Actin patches are highly dynamic structures, in which filaments are being rapidly assembled and disassembled concurrently, and the entire network turns over as rapidly as every 3-4 sec (see above). Here, we discuss the process of branched actin-filament network disassembly, which mechanistically can be broken down into three steps, discussed below: debranching, severing and depolymerization, and monomer recycling.
Quantitative studies in S. cerevisiae and S. pombe suggest that each actin patch contains approximately 5000–7000 actin subunits and 100–300 Arp2/3 complexes and capping proteins (Lin et al. 2010; Sirotkin et al. 2010). Based on these numbers, the actin networks at endocytic sites are predicted to consist of very short filaments with frequent branch points (one branch per 35–50 actin subunits), consistent with EM studies (Young et al. 2004; Rodal et al. 2005a). The Arp2/3 complex–actin filament junction is a specialized structure with a unique F-actin conformation (Rouiller et al. 2008) and thus may require specialized mechanisms for inducing its destabilization. The branches generated by purified actin and yeast Arp2/3 complex are extremely stable for at least 10 min (Gandhi et al. 2010b), whereas in vivo they may be turned over every 3–4 sec. How is this achieved? One factor that influences debranching is the nucleotide state of the Arp2 and Arp3 subunits in Arp2/3 complex. ATP binding (but not hydrolysis) on Arp2 and Arp3 is required for nucleation, and then hydrolysis is thought to occur right after nucleation, which influences debranching (Martin et al. 2005, 2006; Ingerman et al. 2013). Importantly, hydrolysis is not sufficient to promote debranching; additional factors are required. As mentioned above, Gmf1 is a dedicated debranching factor that targets ADP-Arp2/3–actin junctions (Gandhi et al. 2010b; Boczkowska et al. 2013; Ydenberg et al. 2013), and cofilin catalyzes debranching through its ability to bind the sides of ADP-F-actin (Chan et al. 2009). Whether Gmf1 and Cof1 biochemically cooperate in debranching has not yet been tested, but would be consistent with their genetic interactions (Gandhi et al. 2010b; Nakano et al. 2010). A third factor that may contribute to debranching is Crn1, which like Gmf1 binds Arp2/3 complex (Figure 7) and shifts the conformational distribution of the Arp2/3 complex toward the open (inactive) state (Humphries et al. 2002; Rodal et al. 2005b; Ydenberg et al. 2013). These similarities between Crn1 and Gmf1, along with the suggestion that mammalian Coronin-1B promotes debranching (Cai et al. 2008), make Crn1 a promising candidate for participating in pruning daughter filaments.
Filament severing and disassembly:
How is the filamentous content of an actin patch disassembled and turned over in seconds? As mentioned above, the nucleotide state of actin is paramount in controlling turnover, and the conversion of newer (ATP or ADP + Pi) actin to aged (ADP) actin determines what regions of a network are targeted for demolition. At the center of the filament disassembly process is cofilin. COF1 is an essential gene and is required in vivo for the rapid turnover of actin patches (Lappalainen and Drubin 1997). However, an important but often overlooked fact is that cofilin, while essential, is not sufficient to support rates of actin turnover necessary for cell viability. This is evident from genetic observations defining lethal pairs of other actin disassembly factor mutants, including srv2∆ and aip1∆ (Balcer et al. 2003). As described above, there is now strong evidence for at least four other highly conserved actin-binding proteins (coronin, Srv2, Aip1, and Gmf1), and possibly Twf1, functioning in concert with cofilin to catalyze severing and disassembly through a complex multicomponent mechanism (Figure 7). Thus, cofilin should be viewed as one essential component of a larger ensemble that decorates filaments and induces rapid severing and disassembly. At actin patches, the products of enhanced severing are envisioned to be very short oligomers of F-actin, blocked at their barbed ends from further monomer addition either by physical obstruction of the end or conformational changes in F-actin mediated by disassembly factors. These short F-actin fragments would depolymerize into ADP-monomers (many of which should be cofilin bound) as a result of pointed end dissociation. This step is inherently slow in purified F-actin, so it will be important to determine whether it can be accelerated by one or more of the disassembly factors (Figure 7).
The process of actin-filament disassembly produces a rapid and continuous flow of cofilin-bound ADP-actin monomers, and possibly free ADP-actin monomers, which must be rapidly converted back to ATP-actin to enable new rounds of filament growth. The free ADP-actin monomers can be readily recycled due to the inherently fast rate of nucleotide exchange of G-actin, and this step can be accelerated further by binding of profilin or Srv2 (Balcer et al. 2003; Chaudhry et al. 2010). However, the cofilin-bound ADP-actin monomers pose a much greater barrier to recycling for multiple reasons. First, cofilin strongly suppresses nucleotide exchange. Second, cofilin and profilin competitively bind G-actin. Third, cofilin binds with at least 20-fold higher affinity than profilin to ADP-G-actin (Eads et al. 1998; Blanchoin and Pollard 1999). Indeed, profilin has minimal effects in promoting nucleotide exchange on cofilin-bound ADP-actin in vitro (Chaudhry et al. 2010). In addition, plant profilins fail to catalyze nucleotide exchange on actin even in the absence of cofilin (Chaudhry et al. 2007), and yeast profilin has much weaker nucleotide exchange activity than mammalian profilin. In contrast, Srv2 exhibits a very high affinity for ADP-G-actin (20 nM) and readily displaces cofilin from ADP-actin monomers to efficiently catalyze the conversion of cofilin-bound ADP-actin to ATP-actin. Further, it has a binding site for profilin, which may enable profilin and Srv2 to work together in promoting nucleotide exchange (Figure 4) and/or facilitate loading of profilin onto ATP-actin monomers to more rapidly replenish the assembly-competent monomer pool. More work is needed to determine the molecular mechanisms by which these, and possibly other factors, work in concert to control rapid recycling of actin monomers.
Vesicle Scission and Uncoating
The process of remodeling lipid membranes for invagination and scission at endocytic sites is intimately tied to lipid composition and dynamic lipid modifications. Changes in membrane shape rely on force generation by the actin cytoskeleton and are mediated by physical connections between the membrane, coat, and actin network. The role of the membrane and the functions of interacting machinery during endocytosis is an area of intense current investigation, likely to yield many new insights in the coming years. Below we summarize what is known about the roles played by specific lipids in this process. The interested reader is referred to recent literature that provides a more in-depth discussion of the currently understood roles of lipids and lipid-binding proteins that are hypothesized to mediate vesicle scission (Kishimoto et al. 2011; Kukulski et al. 2012a; Smaczynska-De et al. 2012).
The uncoating process is dominated by protein kinases that phosphorylate the coat proteins, disrupt their interactions, and thus promote disassembly. As summarized below, uncoating is necessary to recycle the coat proteins for subsequent rounds of endocytosis, as well as to create coat-free zones on the vesicle surface and thus allow contact and fusion with endosomes.
Roles for lipids and vesicle scission in endocytosis
Sphingolipids and sterols:
The first link between endocytosis and sphingolipids in yeast came from the discovery that a mutation in the long-chain base biosynthetic enzyme Lcb1 caused endocytic defects (Munn and Riezman 1994). Subsequently, it was shown that Pkh1/2 (sphingolipid-stimulated protein kinases) and Ypk1/2 (Pkh1/2-activated protein kinases) also contribute to the regulation of endocytosis (Friant et al. 2001; Dehart et al. 2002). The Sur7 family of proteins is not only implicated in sphingolipid metabolism (Young et al. 2002), but Sur7 was initially found as a suppressor of rvs167 (Sivadon et al. 1997). Interestingly, Rvs167 is the yeast endophilin/amphiphysin homolog and has been directly linked to the scission of invaginated membrane tubules to generate endocytic vesicles (see below). More recently, several endocytic proteins including Rvs167 were shown to bind sphingolipids in a high-throughput lipid-binding screen (Gallego et al. 2010).
Sphingolipids and ergosterol interact to form subdomains in the plasma membrane of yeast and are genetically and functionally linked (Guan et al. 2009). Ergosterol is the yeast version of cholesterol. The plasma membrane has a 3:1 ratio of ergosterol to phospholipids (Zinser et al. 1991); this high level of ergosterol leads to low fluidity/high rigidity of the plasma membrane, restricting the diffusion of transmembrane proteins (Valdez-Taubas and Pelham 2003). A combination of endocytosis and recycling leads to the polarized localization of some transmembrane proteins, such as the v-SNARE Snc1 (Valdez-Taubas and Pelham 2003). Recently, it has become clear that ergosterol and its biosynthetic enzymes are clustered in plasma membrane subdomains referred to as membrane compartment of Can1 (MCCs). When these membrane domains are disrupted, there are global consequences on the structure and organization of the plasma membrane; however, it is currently disputed whether this leads to elevated rates of endocytosis of some cargos or has no effect (Grossmann et al. 2008; Brach et al. 2011).
Ergosterol was first linked to endocytic functions by the Riezman lab’s identification of a mutation in ERG2/END11 in a genetic screen for endocytosis mutants (Munn and Riezman 1994). Ergosterol biosynthetic enzymes that generate a specific sterol ring structure are required for hyperphosphorylation and internalization of Ste2, presumably via the Yck1/2 protein kinases (Munn et al. 1999). However, the deletion of several other ERG genes had no effect on internalization of fluid phase or membrane markers and instead slowed their rate of transport through the intracellular endosome/vacuole system (Heese-Peck et al. 2002).
Phosphoinositides are low abundance, highly bioactive lipids in the cytosolic leaflet of cellular membranes. Different hydroxyl groups on the inositol head group of phosphatidylinositol (PtdIns) are phosphorylated by specific lipid kinases at particular organelles. This differential phosphorylation can convey information that, along with sorting signals in cargo, recruits adaptors to initiate formation of a transport vesicle at the appropriate compartment. Lipid phosphatases counteract the kinase activities and promote dissociation of adaptors.
PtdIns(4,5)P2 is the most abundant phosphoinositide in the plasma membrane, and early clues that its metabolism might regulate the endocytic machinery came from studies of mammalian systems (Beck and Keen 1991; McPherson et al. 1996). Subsequently, lipid kinases and phosphatases required for the biosynthesis and metabolism of PtdIns(4,5)P2 were also found to contribute to endocytosis in yeast cells (Singer-Kruger et al. 1998; Audhya et al. 2000). In endocytosis, PtdIns(4,5)P2 is thought to help localize endocytic machinery to the initiating endocytic site at the plasma membrane. PtdIns(4,5)P2 degradation by the yeast synaptojanin-like Inp51/52/53 enzymes in turn mediates the release of that machinery after vesicle scission. However, the altered dynamic behavior of endocytic components in inp51∆ inp52∆ cells suggests that PtdIns(4,5)P2 turnover regulates multiple stages of endocytosis and not just the disassembly stage (Sun et al. 2007). As mentioned above, PtdIns(4,5)P2 disrupts yeast profilin-actin interactions (Lassing and Lindberg 1985; Goldschmidt-Clermont et al. 1991; Ostrander et al. 1995) and in other systems weakens interaction between CP and actin-filament ends (Schafer et al. 1996). In addition, recent evidence suggests that PtdIns(4,5)P2 can relieve Myo5 autoinhibition, leading to Arp2/3-mediated actin nucleation (Fernández-Golbano et al. 2014). Thus, phosphoinositides in the membrane may directly govern the activities of multiple components of the actin cytoskeleton during endocytosis.
The subcellular localization of specific phosphoinositide species such as PtdIns(4,5)P2 is studied using tools such as GFP–PH–PLCδ (Varnai and Balla 2008). Interestingly, the distribution of this reporter in the plasma membrane originally appeared to be rather uniform; however, more recent TIRFM/2D-deconvolution shows that PtdIns(4,5)P2 is actually found in subdomains of the plasma membrane (Stefan et al. 2002; Spira et al. 2012), suggesting that specific mechanisms govern the synthesis, degradation, or sequestration of PtdIns(4,5)P2 (Ling et al. 2012).
Another intriguing aspect of lipids in promoting endocytosis comes from studies of phospholipid flippases, which facilitate the movement of lipid molecules between the two membrane leaflets (Chen et al. 1999; Farge et al. 1999). Drs2 and related flippases have been indirectly linked to endocytosis; the Drs2 chaperone Cdc50 is a high-copy suppressor of a temperature-sensitive myo3 myo5 double mutant (Gall et al. 2002; Misu et al. 2003), and Drs2 depends upon Pan1 for its correct subcellular localization (Liu et al. 2007). Perhaps most intriguingly, phospholipid flippases may contribute to membrane curvature and budding by promoting bilayer asymmetry (Graham 2004). This is reminiscent of recent discoveries in ER transport vesicle formation, where highly asymmetric cargo was found to promote membrane curvature (Copic et al. 2012). Such a mechanism may explain how budding occurs in cells lacking some bona fide membrane remodeling proteins, such as Rvs161 and Rvs167 (see below), by providing an alternative force to drive membrane deformation and vesicle scission (Kishimoto et al. 2011).
Proteins that promote vesicle scission:
A significant open question is the mechanism(s) by which scission of a membrane bilayer occurs, liberating a vesicle into the cytosol. In animal cells, the key factors implicated in scission are the GTPase dynamin and the BAR domain proteins amphiphysin and endophilin (Campelo and Malhotra 2012). The yeast homologs of amphiphysin and endophilin are Rvs161 and Rvs167, which form a heterodimer (Sivadon et al. 1995; Friesen et al. 2006). As described above for the F-BAR domain of Syp1, BAR domains are dimeric curved structures with a positively charged concave surface that binds to curved membranes (Peter et al. 2004). Some BAR domains also have the capacity to induce curvature by inserting either a helix or a wedge of protein into one leaflet of a membrane bilayer. The Rvs161/167 heterodimer was first implicated in scission when it was observed in rvs mutants that ∼20% of the cortical patches that had begun to move into the cytosol then moved back to the plasma membrane instead of separating from the membrane and being internalized as a vesicle; this was referred to as a retraction phenotype (Kaksonen et al. 2005). Further, the Rvs proteins localize to the midpoint of the invagination, consistent with a function at the position where scission occurs.
A second endocytic vesicle scission mechanism that is well studied in animals involves the membrane-binding GTPase dynamin. Dynamin has an oligomerization-stimulated GTPase activity, a membrane-targeting pleckstrin homology domain, and an SH3 domain that recruits a wide range of PRD-containing proteins (Ferguson and De Camilli 2012). The dynamin oligomer undergoes nucleotide-dependent conformational changes that lead to the scission of the oligomer-wrapped membrane tubule. The closest yeast homolog of dynamin is Vps1, which is necessary for vesicle formation at the Golgi (Vater et al. 1992), binds to Rvs167 and is reported to have a retraction phenotype, suggesting that it facilitates vesicle scission at the plasma membrane during endocytosis (Yu and Cai 2004; Smaczynska-De et al. 2010, 2012).
Other recent work has suggested a requirement for actin forces and lipid modifications in scission, since the myosin I proteins Myo3/5 and the PtdIns(4,5)-5-phosphatase Inp52/Sjl2 support the scission function of the Rvs proteins (Kishimoto et al. 2011). Sla2, interacting with F-actin, also cooperates with the scission machinery (Boettner et al. 2011). Thus, while the magnitudes of the contributions of actin, dynamin, and other conserved components to endocytosis may differ between yeast and metazoans, fundamentally the same machinery and mechanisms are utilized.
Uncoating: Regulation of endocytic proteins through phosphorylation
The final stage of endocytosis is the disassembly of the coat; this allows the components to be reused in new rounds of endocytosis and exposes the membrane of the vesicle for fusion with an endosome. The conserved Ark1/Prk1 kinases have been linked to this disassembly activity (Sekiya-Kawasaki et al. 2003), and indeed they appear at endocytic sites during the later stages, when Abp1 and F-actin appear (Zeng et al. 2007; Toret et al. 2008). Intriguingly, mammalian cells express related protein kinases that also phosphorylate endocytic components to regulate their activities (Greener et al. 2000; Umeda et al. 2000; Conner and Schmid 2002).
Prk1 was first discovered as a spontaneous suppressor of a pan1 allele, and Ark1 was identified as a binding partner of Sla2 (Cope et al. 1999; Zeng and Cai 1999). The Ark1 and Prk1 kinases appear to be largely redundant in their functions, although there is evidence that they also fulfill some unique roles (Cope et al. 1999; Watson et al. 2001; Aghamohammadzadeh et al. 2014). A third related protein kinase, Akl1, is much less well studied. Many endocytic proteins contain multiple copies of the Prk1 consensus phosphorylation sequence [L/I/V/M]xx[Q/N/T/S]xTG, where there is an absolute requirement for the bolded threonine as the acceptor residue (Zeng and Cai 1999; Zeng et al. 2001; Huang et al. 2003). In ark1∆ prk1∆ double-mutant cells, or in cells where the kinase activity is acutely inhibited, large aggregates of membrane, endocytic proteins, and F-actin rapidly accumulate in the cytoplasm (Watson et al. 2001; Sekiya-Kawasaki et al. 2003). These protein kinases disrupt the activities of and/or interactions of specific endocytic components (Watson et al. 2001; Zeng et al. 2001; Chang et al. 2002), e.g., interactions between Pan1 and Sla1, but not Pan1 and End3 (Zeng et al. 2001; Toshima et al. 2007; Chi et al. 2012). Phosphorylation of Pan1 is also suggested to influence its ability to interact productively with F-actin and Arp2/3 complex (Toshima et al. 2005). Two other factors that affect disassembly of the Pan1 complex are the Arf3 GTPase-activating proteins Gts1 and Lsb5, both of which arrive late to the endocytic coat (Toret et al. 2008). Finally, phosphorylation may play a role in uncoating clathrin, since the yeast auxillin homolog Swa2 implicated in this process is heavily phosphorylated (Gall et al. 2000; Krantz et al. 2013).
Phosphorylation of endocytic proteins is counteracted by protein phosphatases, which allow complexes to reassociate and thus participate in new rounds of endocytosis. There is substantial evidence that dephosphorylation of many endocytic proteins is mediated by Scd5, which serves as an adaptor for the PP1 protein phosphatase Glc7 (Chang et al. 2002; Zeng et al. 2007). Sla1 contains a Glc7 binding site that is required for efficient endocytic function and turnover of phosphorylated Sla1. Pan1 and Ent1 also have predicted Glc7 binding sites (Gardiner et al. 2007). Recent work shows that Pan1 dephosphorylation and disappearance from patches depends on Pan1 interactions with End3 and recruitment of Scd5 (Whitworth et al. 2014). Thus, there is a critical role for the Scd5–Glc7 complex in regulating the dynamics and behavior of the endocytic machinery at cortical patches (Chi et al. 2012).
The results above highlight the importance of phosphoregulation in endocytosis, particularly in the final uncoating stage (Sekiya-Kawasaki et al. 2003). Such regulation also facilitates crosstalk between the dynamic protein machinery and the underlying lipid bilayer, e.g., through the PtdIns(4,5)-5-phosphatase Inp52/Sjl2 and the membrane-binding endocytic factors Abp1 and Rvs167 by the kinase Tor2 (Kishimoto et al. 2011; Tenay et al. 2013). Further studies are needed to map the phosphorylation states of other components of the endocytic machinery and thus understand the full extent of phosphoregulation in endocytosis.
Latest Findings, Emerging Views, and Open Questions
Temporal control of endocytosis
Ultrastructure of endocytic sites over time:
As with many cellular structures that are below the resolution limit of standard light microscopy, it has been challenging to assign precise locations to proteins at endocytic sites and to determine how their spatial patterns change during the rapid steps of membrane invagination and vesicle scission. The two most fruitful approaches have been quantitative immuno-EM and correlative light and electron microscopy (CLEM) (Idrissi et al. 2008; Kukulski et al. 2011; Idrissi et al. 2012). One recent CLEM study concluded that membrane bending begins with actin polymerization (Kukulski et al. 2012a); this seems reasonable given that substantial membrane bending requires forces from actin polymerization, especially given the combination of high turgor pressure and a stiff plasma membrane rich in ergosterol (Basu et al. 2014; Carlsson and Bayly 2014). However, a different study using immuno-EM reported that membrane bending preceded the arrival of actin (Idrissi et al. 2012). These differences highlight a need for continued efforts on this problem.
Another unresolved issue is the orientation of the actin filaments in the developing patch, which requires having ultrastructural information. To date, it has been extremely challenging to perform myosin S1 decoration coupled with EM to define the orientation of filaments in yeast actin structures. However, Mabuchi and colleagues managed to obtain ultrastructures of S. pombe actin cables and cytokinetic rings (Kamasaki et al. 2007; Mishra et al. 2013), suggesting that with more effort it may be possible to do the same for cortical patches. In the absence of such information for actin patches, models of actin-filament architecture assume that the positions of the NPFs and Arp2/3 complex inform the location of the growing barbed ends. One immuno-EM study reported a marked enrichment of Arp2/3 complex compared to other patch components away from the membrane, which was interpreted to suggest that the growing barbed ends of actin filaments face the base of the invagination (Rodal et al. 2005b). Consistent with this view, Las17–GFP was initially reported to localize to the base of the invagination and not move inward with other patch components (Kaksonen et al. 2003). However, more recent studies, one using a Las17–GFP fusion with the tag in an alternative position and one using immuno-EM, concluded that Las17 does move in with the membrane invagination (Galletta et al. 2008; Idrissi et al. 2008). Thus, the field has not yet reached consensus on the locations of NPFs and hence the orientation of F-actin in the cortical patch. More definitive answers may come when higher-resolution instruments are developed and applied to a carefully vetted set of reagents. Further, we anticipate the need to consider whether some of the reported phenotypes of patch mutants have been partially misinterpreted because deletion or tagging of that component caused (then unrealized) compensatory changes, such as premature arrival or departure of another component.
Do checkpoints control transitions Between endocytic coats?:
Multicolor live imaging of patch components in yeast cells has become an incredibly powerful assay for defining the roles and dynamics of different endocytic proteins and the consequences of mutations or perturbations. And from such approaches, it is now clear that dozens of endocytic proteins are recruited in an ordered sequence, organized as coats. However, what remain unclear are the rules governing the recruitment and progression through each step in this coordinated program. The transition times between some stages are quite regular. For instance, in wild-type cells the gap between the arrival of the middle and late coat is typically very short, although in certain mutants the gap is prolonged and more variable (e.g., Newpher and Lemmon 2006; Chi et al. 2012). In contrast, the early-to-middle coat transition is highly variable even in wild-type cells. The source of this variability is not currently known, but may result from a dependence on an increasing local concentration of transmembrane cargo and/or PtdIns(4,5)P2. If so, how concentrated must the cargo or lipid become, and how is this sensed? Since the length of time between initiation and the arrival of the middle coat is the most variable of all the endocytic stages, it seems likely to involve some kind of checkpoint. What mechanism could be at work is pure speculation at this point, but thoughts turn to there being a critical level of cargo, or a critical level of cargo interactions with early coat proteins leading to formation of an organized structure with multiple copies of several different proteins.
Unresolved issues in actin network formation and dissolution
There are also a number of unresolved questions about how actin networks assemble at sites of endocytosis, how the filaments are organized, and how the networks facilitate membrane remodeling. While we know that Arp2/3 complex is the actin nucleator and generates short, branched filaments, we do not know whether all, or even most filaments in a patch are branched. Branches could be rapidly pruned by GMF and cofilin, such that only a fraction of filaments in a patch would be branched at any given point. Further, debranching may occur more robustly in specific regions of a patch, or at specific phases of patch maturation. This remodeling may in turn permit crosslinking proteins such as Sac6 and Scp1 to organize filaments into parallel or orthogonal arrays, establishing heterogenous filament geometries within a patch. Whatever form the F-actin takes, another open question is whether proteins are free to diffuse within the network, or is the internal environment too dense, making only the outermost shell accessible to cytosolic factors?
Another key question is what restricts the size of the actin network. There is a large enough supply of building blocks (actin, Arp2/3 complex, etc.) in the cytosol to build many actin patches at once, so what factors limit the growth of an individual endocytic site? Is there limiting access to membrane or to active NPFs? A valuable clue in further investigating this process is the observation that end3 and sla1 mutants accumulate fewer but aberrantly large patches (Holtzman et al. 1993; Benedetti et al. 1994). A related question is what controls the orientation of polymerization to direct the forces that drive membrane invagination and vesicle scission. Critical to this issue may be the precise locations where each NPF is anchored and/or activated. In this regard, superresolution light microscopy may be very helpful in determining if patches have subdomains enriched or depleted of specific proteins.
There are also important questions left to answer concerning the mechanisms regulating the onset of actin patch assembly, as well as transport and subsequent disassembly of the internalized vesicle/patch. As discussed earlier, it is still unresolved how the first actin filaments are nucleated at an actin patch, i.e., whether there is machinery that activates the Arp2/3 complex for nucleation in the absence of a mother filament, or instead a short mother filament is captured, or if specialized machinery assembles a short “starter” seed (as depicted in Figure 6). Further, how is Las17/WASP activated at the onset of actin network formation? Is Bzz1 the switch, or is additional machinery involved? How is Bzz1 recruited to patches at this time, and does it require its F-BAR domain and/or recognition of specific lipids or specific curvature in the membrane? Once vesicles are internalized, how many of them are transported on actin cables, and how are they attached to cables? If a patch moving inward on a cable does not encounter an endosome, does it travel to the back of the mother cell, and if so, what is its fate? Finally, by what mechanisms do internalized vesicles shed their actin coat, and is this a prerequisite for fusion with endosomes?
Alternative pathways for endocytic internalization
Exciting recent evidence demonstrates the existence of an endocytic route that is independent of the canonical clathrin- and Arp2/3-dependent pathway, and instead depends upon Rho1, the formin Bni1, and tropomyosin (Tpm1), critical factors in actin cable assembly (Prosser et al. 2011). Consistent with this finding, evidence for an Arp2/3-independent endocytic route has also been obtained in the pathogenic yeast Candida albicans (Epp et al. 2010). Still another recent study has described an alternative endocytic pathway that is evident even after pharmacological interference with actin polymerization by Latrunculin A (Aghamohammadzadeh et al. 2014). This pathway was found to be dependent on Abp1 and its binding partner Ark1, but not the related Prk1 kinase, or several other proteins involved in “canonical” CME. Given the low dose of Latrunculin A used in this study though, and the finding that the Arp2/3-binding A region of Abp1 was required, actin dynamics may still play a role in this pathway. Many questions about these two noncanonical endocytic pathways remain unanswered, including whether they are harnessed to specific cargo or activated by specific environmental conditions. Applying many of the same techniques that have shed so much light on the CME pathway (e.g., live cell imaging of GFP-tagged components and electron microscopy) should allow for a more mechanistic understanding of these alternate pathways.
Comparisons between yeast and mammalian endocytic mechanisms
A fundamental distinction between CME in yeast vs. mammalian cells has been the relative functional importance of clathrin and actin in these two systems. Internalization of CME sites in yeast proceeds even in the absence of clathrin, albeit with significantly reduced efficiency (Kaksonen et al. 2005; Newpher and Lemmon 2006), whereas clathrin depletion in mammalian cells arrests CME completely (Motley et al. 2003; Huang et al. 2004). Hence, the current view is that clathrin fulfills a primarily scaffolding/organizational role in yeast, while it mediates a more structural role (in shaping the vesicle) in mammalian cells. In yeast, the requirement for actin polymerization is essential and is believed to contribute the structural role by providing the force for membrane remodeling that opposes the high turgor pressure that exists across the plasma membrane (Aghamohammadzadeh and Ayscough 2009; Basu et al. 2014; Carlsson and Bayly 2014). Further, it has traditionally been thought that there is only a minimal requirement for actin during CME in animal cells. However, recent findings show that the requirement for actin polymerization in animal cell CME can depend on the surrounding environment (Basu et al. 2014). Two types of clathrin structures have been observed in mammalian cells. One is the classic clathrin-coated pit, which forms a spherical clathrin cage structure with an enclosed membrane, leading to internalization even when actin polymerization is blocked (Ehrlich et al. 2004; Loerke et al. 2009). The second is referred to as a plaque and is flat and larger than a traditional clathrin pit; these plaques become associated with F-actin as they are internalized (Saffarian et al. 2009). Actin appears to convert these clathrin scaffolding plaques into internalizing structures, often at locations in cells of high transmembrane pressure or tight contact with other cells or extracellular matrix (Saffarian et al. 2009; Collins et al. 2011). The extracellular matrix is reminiscent of the yeast cell wall, as is the fact that clathrin is not predicted to form a full spherical cage at yeast endocytic sites based on quantification of clathrin abundance per actin patch in S. pombe (Berro et al. 2010; Sirotkin et al. 2010). These observations raise interesting possibilities that require further investigation, including whether there is an analogy to draw between the features of clathrin structures in yeast and the clathrin plaques in mammalian cells. It will be interesting to learn whether the loss of clathrin can be partially tolerated at these sites in mammalian cells as it can be in yeast.
Homeostatic regulation of endocytosis
Response to changing environmental conditions:
Localization and trafficking of many nutrient permeases and other transmembrane proteins in the plasma membrane are regulated by the concentrations of particular compounds in the media. For example, there tend to be both high- and low-affinity permeases for specific molecules (e.g., methionine, tryptophan, inositol, or uracil), and typically the high-affinity permease is rapidly internalized and degraded when the cells are exposed to high concentrations of the substance (Levy et al. 2011). Similarly, when the metabolic state of the cell changes and cytosolic concentrations of certain compounds are affected (e.g., amino acids or sugars), this can again lead to changes in the residence time and trafficking of particular permeases or transporters (Rubio-Texeira and Kaiser 2006). The general concept of a plasma membrane quality-control system has been proposed, i.e., that there is machinery that detects and removes damaged or misfolded proteins from the plasma membrane, although there is some disagreement on the subject (Lewis and Pelham 2009; Zhao et al. 2013). This suggested plasma membrane quality-control system appears to require the activities of the E3 Ub-ligase Rsp5, much like nutrient-triggered transporter internalization (Cargo sorting signals and adaptors), raising the possibility that an intermediate conformation in transporter endocytosis resembles a misfolded protein (Keener and Babst 2013). However, there are some differences in terms of what components work together with Rsp5 to mediate recognition and modification of the substrate, including a recently discovered N-terminal domain in many permeases that directs an apparently novel Rsp5-dependent modification (Keener and Babst 2013).
Response to physiological stress:
The normal polarized localization of endocytic actin patches to the bud is also sensitive to stress, e.g., elevated temperature or high salt. In stressed wild-type cells, the polarization is temporarily lost; patches become distributed over the mother and bud surface within 30 min of the shock, but they regain a polarized distribution within 90–120 min (Desrivières et al. 1998). A common feature of cells with endocytic defects and depolarized actin patches is hyperactivity of cell-wall stress pathways, leading to overly thick cell walls (Tang et al. 2000; Dehart et al. 2003; Ge et al. 2005); this supports the view that the endocytic machinery might internalize enzymes involved in cell-wall remodeling (Chapa-Y-Lazo et al. 2014). A number of factors have been implicated in regulating patch redistribution or depolarization, including the cell-wall stress PKC1–MAPK pathway and perturbed sphingolipid metabolism (Delley and Hall 1999; Balguerie et al. 2002). For detailed insights into the causes of this depolarization, readers are referred to the article by Howell and Lew in this collection of reviews (Howell and Lew 2012).
Cell cycle and polarity:
Cortical actin/endocytic patches have a characteristic distribution pattern in yeast cells, which changes in different stages of mitotic growth (Adams and Pringle 1984; Amberg 1998), suggesting that it is cell-cycle regulated. These distribution patterns are under intense study and are likely to be governed by a combination of inputs that include bud site landmarks, regulatory GTPases that read the landmarks, and other factors such as plasma membrane phospholipids, delivery of secretory cargo by fusion of exocytic vesicles, and feedback loops of signaling pathways (Li et al. 1995; Kuo et al. 2014). Interestingly, this may also involve the morphogenetic checkpoint mediated by Swe1 kinase, which arrests mitosis in response to defects in actin organization and bud growth (McMillan et al. 1998). Recent work in both C. albicans and S. cerevisiae found evidence for Swe1-dependent phenotypes in endocytosis mutants; combining deletion of SWE1 with a sla1 or sla2 mutation led to an increase in multinucleated cells, indicative of a Swe1-dependent checkpoint acting on (and regulating?) components of the endocytic machinery (Gale et al. 2009). Finally, Tor2, a member of the cell-cycle-dependent kinase complex TORC2, appears to play a role in controlling the recruitment of endocytic factors Abp1 and Rvs167 to facilitate internalization of α-factor receptor (Dehart et al. 2003; Tenay et al. 2013).
Nuclear functions of endocytic proteins
It has long been known that several endocytic proteins in mammalian systems undergo nucleocytoplasmic shuttling and can influence transcription (Vecchi et al. 2001). Thus, it was intriguing to find that several yeast endocytic proteins (Scd5, Pan1, Rsp5, and Sla1) likewise spend part of their time in the nucleus (Chang et al. 2006). These studies suggest that endocytic proteins may have other important roles in regulating the assembly and activity of multisubunit complexes in the nucleus and have pointed to possible roles for the karyopherins Crm1 and Kap120 in fluid-phase endocytosis. Additionally, the PtdIns(4)P-5-kinase Mss4 is found in both the cytosol and the nucleus, but it must localize to the plasma membrane to generate PtdIns(4,5)P2 and support viability (Audhya and Emr 2003). Bcp1 is a nuclear protein that facilitates Mss4 export from the nucleus to cytosol and restores PtdIns(4,5)P2 synthesis, suggesting a possible regulatory cycle between the nuclear and cytosolic/plasma membrane compartments. Given the nucleocytoplasmic shuttling by other endocytic proteins, this regulatory cycle is likely to control more than just the levels of PtdIns(4,5)P2.
New approaches for studying yeast endocytosis and actin dynamics
The relatively small size of its genome and the availability of extremely powerful genetic tools have made yeast a pioneering model organism for genome-wide high-throughput analyses and have formed the basis for many systems biology approaches. For example, automated techniques allowed for systematic combination of nonlethal mutants into every possible pairwise double mutant followed by phenotypic analysis (Costanzo et al. 2010). This has expanded to analyzing more subtle phenotypes by combining mutations with strains expressing GFP-tagged proteins and determining effects on protein localization or abundance (Breker et al. 2013). The small number of proteins encoded by the yeast genome have also made it a fertile testing ground for identifying new protein complexes through genome-wide purification of TAP-tagged proteins coupled to mass spectrometry or genome-wide yeast two-hybrid screens and other protein–protein interaction strategies. The strength of analyzing the effects of one or a few mutations on the whole organism carries over into the ability to test the predictions of sophisticated computational modeling.
The innovations described above played an instrumental role in bringing about key discoveries over the past 17 years and will undoubtedly continue to serve as critical tools in addressing open questions about actin and endocytosis over the coming decades. However, it is anticipated that the next wave of advances will also rely on development and application of new tools and instrumentation, as well as new strategies and approaches. Super-resolution light microscopy may enable one to more precisely define the locations of patch proteins and subdomains at endocytic sites. The development of new fast-acting chemical inhibitors that target specific endocytic patch proteins may enable one to acutely arrest patches in specific stages of their maturation and provide new mechanistic insights into the functional roles of each patch component. Moreover, it may be possible to reconstitute some of the steps in endocytosis in vitro and to develop single molecule assays to define the activities and interactions of early patch components. Arguably, many of the important advances made in understanding the mechanisms of actin network formation in endocytosis have come from sophisticated biochemical assays and mathematical modeling, which have provided rigorous, quantitative information about each actin regulatory protein. Future advances are likely to require similar assays for the functions of earlier components in the pathway and will enable us to define their order of assembly, affinities, and specific contributions to the process. Ultimately, this anticipated increase in knowledge of the activities, interactions, and regulation of each component of the endocytic patch may enable us to build a complete model for endocytosis and to test experimentally its design principles and gain unprecedented understanding of this biological process.
We are extremely grateful to Kate Bradford, Melissa Cataldo, Kyle Hoban, Derek Prosser, Avital Rodal, Alex Wade, Kristie Wrasman, and Casey Ydenberg for critical reading of the manuscript. We also thank Sandra Lemmon, Amanda Reider, and Avital Rodal for helpful discussions. We apologize to the many researchers in this field whose important contributions we did not discuss due to space limitations. This work was supported in part by National Institutes of Health grants to B.L.G. (GM063691) and B.W. (GM060979).
Available freely online through the author-supported open access option.
Communicating editor: T. N. Davis
- Received August 29, 2014.
- Accepted November 26, 2014.
- Copyright © 2015 by the Genetics Society of America
Available freely online through the author-supported open access option.