Clathrin-mediated endocytosis is an essential cellular process that involves the concerted assembly and disassembly of many different proteins at the plasma membrane. In yeast, live-cell imaging has shown that the spatiotemporal dynamics of these proteins is highly stereotypical. Recent work has focused on determining how the timing and functions of endocytic proteins are regulated. In this Cell Science at a Glance article and accompanying poster, we review our current knowledge of the timeline of endocytic site maturation and discuss recent works focusing on how phosphorylation, ubiquitylation and lipids regulate various aspects of the process.
Clathrin-mediated endocytosis (CME) is a highly conserved essential cellular process by which external materials, integral membrane proteins and membrane phospholipids are internalized into the cell. The internalization of these materials is accomplished through deformation of the plasma membrane by a two-dimensional protein lattice, comprising clathrin and associated proteins. The canonical model in both yeast and mammalian cells is that the clathrin lattice and the underlying plasma membrane invaginate to form clathrin-coated pits (CCP), which eventually pinch off from the plasma membrane as clathrin-coated vesicles (CCV) (Kirchhausen et al., 2014). The CCV is then uncoated, and the resulting vesicle fuses with early endosomes, where cargo is sorted for degradation or recycling back to the plasma membrane (Grant and Donaldson, 2009). CME is especially important for the regulation of ligand-mediated receptor tyrosine kinase signaling pathways, such as epidermal growth factor receptor (EGFR) and insulin receptor (INSR) signaling; here, CME functions to internalize ligand-bound receptors from the plasma membrane for their routing to the endosomal membrane systems (reviewed in Di Fiore and von Zastrow, 2014; Goh and Sorkin, 2013). Although CME generally maintains cellular homeostasis, it can also be hijacked for entry into cells by certain viruses and bacteria, such as influenza A and Listeria (McMahon and Boucrot, 2011).
The budding yeast Saccharomyces cerevisiae has been an invaluable model organism for studying CME. The major route into the cell is mediated by clathrin, although it is interesting to note that a clathrin-independent endocytosis (CIE) pathway in S. cerevisiae has been found (Prosser et al., 2011) and is involved in a specific form of α-arrestin-regulated receptor internalization (Prosser et al., 2015). Budding yeast is amenable to molecular genetic manipulations, making it possible to genetically tag proteins of interest with fluorescent proteins in order to study their localization and dynamics at endogenous levels. Budding yeast is also advantageous for real-time imaging because the cells are spherical, and individual CME events can be followed from inception at the cortex to internalization into the cytoplasm by imaging from the side using a medial focal plane. Because most CME components are conserved from yeast to humans, principles learned in yeast are expected to apply broadly. For example, Sla2 is a key protein in yeast CME that links the actin cytoskeleton to the coat machinery. In sla2Δ cells, branched actin polymerization is decoupled from the endocytic machinery, resulting in an accumulation of filamentous actin (F-actin) ‘tails’ comprising branched filaments that polymerize at the cortex and flux into the cytoplasm (Kaksonen et al., 2003). When the human homolog, Hip1R, is knocked down, cells are defective to a similar extent in endocytosis and accumulate cortical actin structures (Engqvist-Goldstein et al., 2004). Fluorescent imaging of CME in mammalian cells has shown that endocytic site dynamics are similar to those of yeast CME (Taylor et al., 2011). The advent of genome-editing tools now allows for expression of fluorescently tagged proteins at the endogenous level, similar to what has been done in yeast, and has been used to further show that the process is, overall, very similar in both organisms (Doyon et al., 2011; Hong et al., 2015). Although CME has also been well studied in other fungi, such as the fission yeast Schizosaccharomyces pombe, we chose to focus here on recent highlights from S. cerevisiae for the sake of brevity.
In budding yeast, a detailed timeline of the arrival and departure of over 60 proteins at endocytic sites was established in the early- to mid-2000s (Merrifield and Kaksonen, 2014). These proteins can be organized into modules according to their functions and the timing of their appearance and disappearance at endocytic sites: (i) early proteins; (ii) early, middle and late coat proteins; (iii) WASp (also known as Las17 in yeast) and myosin-related proteins, (iv) actin and actin-associated proteins; and (v) scission-related proteins (see poster). Although much effort has been put into understanding the spatiotemporal dynamics of these modules with high precision, here we review recent studies that have focused on understanding, at a molecular level, what kinds of protein and lipid modification regulate the progression of events at an individual CME site. We specifically focus on protein phosphorylation and ubiquitylation, as well as lipid modifications and protein interactions with lipids. Finally, we highlight new and emerging studies that bridge cargo regulation to internalization.
Almost all endocytic proteins across all of these modules are phosphorylated (see poster). Several kinases and their substrates have been identified at endocytic sites. In addition, there are kinases that have not been found at endocytic sites, but whose activities have effects on endocytosis.
A member of the conserved casein kinase family, Hrr25, most closely related to the human CK1δ and CK1ε isoforms (Petronczki et al., 2006), has recently been identified as a member of the early protein module and found to phosphorylate many CME components (Peng et al., 2015a). Interestingly, Hrr25 is also involved in many other cellular processes, including autophagy (Tanaka et al., 2014), trafficking through the Golgi (Lord et al., 2011), DNA damage repair (Hoekstra et al., 1991) and mitotic spindle function (Peng et al., 2015b; Petronczki et al., 2006). Whether Hrr25 coordinates any of these activities with CME needs to be determined.
Several endocytic proteins are also known to be phosphorylated by members of the Ark kinase family, which, in mammalian cells, includes GAK and AAK1, and in yeast, includes Ark1, Prk1 and Akl1 (reviewed in Smythe and Ayscough, 2003). Ark1 and Prk1 (see the ‘actin’ module on the poster) have well-described roles in controlling actin dynamics at endocytic sites through phosphorylation of Pan1 and End3 (Cope et al., 1999; Toshima et al., 2005; Zeng and Cai, 1999; Zeng et al., 2001), and in disassembling coat proteins after the CCV internalization (Sekiya-Kawasaki et al., 2003; Toret et al., 2008).
Many endocytic proteins are also substrates of kinases that have yet to be detected at endocytic sites. Several, such as Abp1, Sla1 and Sla2, are substrates of Cdk1, the master cell cycle regulator (Holt et al., 2009). The yeast amphiphysin homolog, Rvs167, is phosphorylated by the cyclin-dependent kinase Pho85 and the mitogen-activated protein kinase (MAPK) Fus3 (Friesen et al., 2003; Lee et al., 1998). Phosphorylation of Rvs167 by Pcl2–Pho85 in vitro disrupts a physical interaction with Las17/WASp (Friesen et al., 2003). The type-I myosin Myo5 is regulated by Cka2, which is the catalytic subunit of the casein kinase 2 holoenzyme [CK2; see the section on lipid requirements and interactions below for more details (Fernández-Golbano et al., 2014)]. The yeast AGC kinaseYpk1 and its upstream kinase Pkh1, functional homologs of SGK1 and its upstream kinase 3-phosphoinositide-dependent kinase 1 (PDPK1), respectively (Casamayor et al., 1999), are also required for receptor-mediated and fluid-phase endocytosis; they are of particular interest because they are also involved in sphingolipid signaling (Friant et al., 2001; deHart et al., 2002; see below). Future studies of these kinase-substrate relationships will determine if and how CME responds to other cellular events.
To fully understand how phosphoregulation functions in CME, it is also necessary to understand de-phosphorylation. Scd5, a subunit that targets protein phosphatase 1 (PP1, Glc7 in yeast) (see poster), counteracts phosphorylation by Ark1 and Prk1 (Chang et al., 2002; Chi et al., 2012; Henry et al., 2003; Zeng and Cai, 1999; Zeng et al., 2007). In scd5 mutants that fail to interact with PP1/Glc7, many endocytic adapter proteins in the coat module are hyperphosphorylated (Chi et al., 2012; Zeng et al., 2007). In addition, lack of Scd5 activity causes a delay in the transition from mid to late coat protein progression, which is likely to be due to inefficient recruitment of hyperphosphorylated Sla1 (Chi et al., 2012).
Ubiquitylation of the cytosolic domains of cargo regulates CME internalization in both yeast and mammals (Haglund and Dikic, 2012; Traub and Lukacs, 2007). A long-standing goal is to determine how cargo affects the progression of endocytic site maturation. Ubiquitin is likely to play a role in linking cargo to the CME machinery. Three of the known yeast endocytic proteins are known to have mono-ubiquitin-interacting domains – a ubiquitin-associated (UBA) domain in Ede1, and a ubiquitin-interacting motif (UIM) in both Ent1 and Ent2 (see poster). These proteins regulate the ligand-induced internalization of the α-factor pheromone receptor Ste2, which is regulated both by casein kinase Yck2-mediated phosphorylation and ubiquitylation by the HECT E3 ubiquitin ligase Rsp5 (Dunn and Hicke, 2001a,b; Toshima et al., 2009). Ent1 and Ede1 are also thought to be recruited to endocytic sites by ubiquitylated cargo (Aguilar et al., 2003). Recent studies are also now focusing on a class of regulatory proteins called α-arrestins in the regulation of ubiquitylation and in Rsp5-dependent cargo internalization, which will inform how cargoes are specifically regulated and selected for internalization (Becuwe et al., 2012; Prosser et al., 2015).
In addition to conventional ubiquitin-interacting domains, one of the SH3 domains of Sla1 also binds to ubiquitin (Stamenova et al., 2007). This is intriguing as SH3 domains typically bind to the core peptide motif PxxP (where ‘x’ represents any amino acid), which exists in many endocytic proteins. Interestingly, a PxxP domain that has bound to this particular ubiquitin-binding SH3 domain can be competed off with free ubiquitin (Stamenova et al., 2007). This competition is a potential mechanism for regulating protein–protein interactions in space and time.
Studies have also begun to focus on a role for ubiquitylation of the endocytic machinery itself. Many endocytic proteins, including Sla1 and Rvs167, have been identified as targets of Rsp5 (Stamenova et al., 2004). The early protein Ede1 appears to have multiple levels of regulation through interaction of its UBA domain with ubiquitin, as well as by being ubiquitylated itself (Dores et al., 2010). In yeast, deletion of the deubiquitylases UBP7 and UBP2 – whose gene products antagonize Rsp5-mediated ubiquitylation – or expression of an Ede1–ubiquitin fusion, results in formation of internal punctae, which contain many endocytic proteins and show similar dynamics to bona fide CME sites. This result has been hypothesized to be due to hyper-ubiquitylated Ede1 and misregulated CME-site initiation. Ubp7 has been found to be a component of endocytic sites, and the protein shows similar spatiotemporal temporal dynamics, as well as spatial localization, to other proteins of the ‘WASp and myosin’ module. These proteins do not internalize with the late coat and actin proteins before vesicle scission (see poster) (Weinberg and Drubin, 2014). In addition, the ubiquitin regulatory X (UBX)-domain-containing protein Ubx3 regulates the rate of Ede1 recruitment to endocytic sites (see poster). This regulation is proposed to be mediated by an interaction between Ubx3 and Cdc48, an AAA-ATPase ubiquitin-editing complex (Farrell et al., 2015). These observations suggest that dynamic ubiquitylation and deubiquitylation are important for regulating the role of Ede1 in CME site initiation and maturation. It will be important to test whether there is a functional relationship between Ede1 ubiquitylation and phosphorylation.
Lipid functions and interactions
In almost every module, from early site nucleation to vesicle scission, there is at least one lipid-binding protein (see poster). Lipids have very important roles in the endocytic timeline, yet they are more challenging to study than proteins, somewhat hindering elucidation of their functions. The lipid-binding activities of endocytic proteins might be important to mediate membrane bending or to harness actin forces.
Turnover of phosphatidylinositol-4,5-bisphosphate [PtdIns(4,5)P2] is required for endocytic function (Sun and Drubin, 2013; Sun et al., 2007). Overproduction of PtdIns(4,5)P2 by deleting genes encoding two of the three PtdIns(4,5)P2 phosphatases, the synaptojanins Sjl1 and Sjl2, leads to a delay in the disassembly of the coat and actin, and defects in fluid-phase and receptor-mediated endocytosis, possibly owing to a failure in scission (Singer-Krüger et al., 1998; Sun et al., 2007). Sjl2 itself is recruited to endocytic sites by Abp1 (Stefan et al., 2005) (see poster). Conversely, PtdIns(4,5)P2 depletion using a temperature-sensitive mutant of Mss4, the sole budding yeast phosphatidylinositol-4-phosphate-5-kinase, results in formation of actin tails; in this mutant, coat proteins do not internalize, but actin continuously streams inwards from these coat proteins (Sun and Drubin, 2013). The Epsin N-terminal homology (ENTH) and AP180 N-terminal homology (ANTH) domains both bind to PtdIns(4,5)P2, and are present in the coat proteins Ent1 and Ent2 (intermediate coat), Yap1801 and Yap1802 (early coat), and Sla2 (intermediate coat) (Aguilar et al., 2003; Carroll et al., 2012; Ford et al., 2001; Itoh et al., 2001; Sun et al., 2004). The ENTH and ANTH domains of Ent1 and Sla2 co-assemble in a PtdIns(4,5)P2-dependent manner and have been proposed to be responsible for tethering the endocytic machinery to the plasma membrane during membrane tubulation (Skruzny et al., 2012, 2015).
Whether and how lipids play a role in site initiation is still an open question. A clue comes from cells in which RCY1, whose gene product is involved in endosome-to-Golgi trafficking, is deleted. In rcy1Δ cells, phosphatidylserine, but not PtdIns(4,5)P2, accumulates on abnormal internal membrane compartments. Endocytic proteins accumulate on these compartments before the proteins dissociate in the same temporal manner as observed on the plasma membrane, indicating a possible role for phosphatidylserine, specifically, in CME site initiation (Sun and Drubin, 2013). Interestingly, phosphatidylserine is required for cell polarity and is itself localized in a polarized manner, being enriched on the daughter cell following cell division (Fairn et al., 2011). The daughter cell has been observed to have more endocytic sites than the mother, indicating that phosphatidylserine could have a role in site initiation (Bi and Park, 2012). When components in the phosphatidylserine biosynthetic pathway are mutated, endocytic sites lose their polarized localization (Sun and Drubin, 2013). However, a phosphatidylserine-binding protein specifically located at CME sites has yet to be identified.
The endocytic role for sphingolipids, including sphingoid bases and ceramides, has been elusive. An lcb1 mutant, which is deficient in sphingoid base synthesis, the first step in the synthesis of more complex sphingolipids, shows a substantial defect in internalization of α-factor pheromone, the classic yeast CME cargo (Munn and Riezman, 1994; Zanolari et al., 2000). Further studies are needed to understand the mechanistic basis for this phenotype.
Many lipid-binding proteins, and specifically proteins that either sense or generate curvature, are involved in the final scission process. The yeast amphiphysin homolog, a heterodimeric complex of Rvs161 and Rvs167, is concentrated along the neck of the growing tubule of a deeply invaginated CCP (Idrissi et al., 2008; Kaksonen et al., 2005; Picco et al., 2015). Bzz1, an F-BAR protein of the WASp and myosin module, and Rvs161–Rvs167 have been found to contribute cooperatively to vesicle scission (Kishimoto et al., 2011). Absence of Rvs161–Rvs167 results in a ‘yo-yo’ phenotype, wherein coat and cargo components internalize from the cortex but then retract back to and disassemble at the cortex instead of entering the cell. This phenotype is exacerbated by loss of Bzz1. The lack of Rvs161–Rvs167 and Bzz1 results in an improper geometry of invagination (Kishimoto et al., 2011). Similarly, deletion of VPS1, whose gene product is a dynamin-like protein, results in ‘yo-yo’ phenotypes and misshaped invaginations (Smaczynska-de Rooij et al., 2010), consistent with a role in regulating membrane shape and scission, potentially through an F-actin-bundling activity (Palmer et al., 2015). Unlike the conventional dynamins involved in mammalian CME, and similar to other dynamin-like proteins, Vps1 does not contain a pleckstrin-homology (PH) domain. However, it can still bind to and tubulate liposomes (Smaczynska-de Rooij et al., 2010). Although Vps1 does not appear to play as central of a role in scission as dynamin does in mammalian cells, it, along with several other lipid-binding proteins, such as Bzz1 and the Rvs161–Rvs167 complex, might coordinate membrane remodeling and vesicle scission.
Many studies have focused on how the forces and geometries needed to generate membrane scission can be generated by actin and the late-arriving scission proteins (Kaksonen et al., 2003, 2005; Picco et al., 2015; and reviewed in Goode et al., 2015; Idrissi and Geli, 2014). Nevertheless, whether the membrane bends before or after actin arrives is still debated (see poster; Idrissi et al., 2012; Kukulski et al., 2012). Therefore, determining the geometry of the endocytic membrane and actin network at each stage is an important goal. In yeast, the two type-I myosins Myo3 and Myo5 have essential lipid-binding and motor activities (Lewellyn et al., 2015). Fluorescence microscopy analyses have revealed that these myosins are restricted to the base of growing pits (Jonsdottir and Li, 2004; Picco et al., 2015). However, immunogold labeling and electron microscopy analyses have demonstrated that the myosins, as well as Las17/WASp, can also be found at the tip of invaginating pits (Idrissi et al., 2012). Accurate localization of these proteins is particularly important for understanding how the geometry and orientation of actin polymerization and myosin lipid-binding activities contribute to the forces that are required for membrane bending during invagination. For example, spatially regulated phosphorylation and repression of Myo5-mediated actin polymerization by Cka2 is proposed to concentrate branched actin polymerization specifically at the base of endocytic sites (Fernández-Golbano et al., 2014). Interestingly, Syp1, an early-arriving protein, is an F-BAR protein that can interact with actin and negatively regulate actin polymerization induced by the much later arriving protein Las17/WASp in vitro (Boettner et al., 2009) (see poster). How this interaction functions in vivo remains an open question.
Tying everything together
The assembly and disassembly dynamics of many CME proteins at endocytic sites have been determined; however, understanding how spatiotemporal regulation is achieved is necessary for understanding the underlying mechanisms for this well-conserved process. Based on the intriguing observation that early-arriving proteins have highly variable lifetimes compared to the very regular timing of later-arriving proteins (late coat and beyond), CME can be broadly simplified into two temporal phases – an ‘early phase’ and a ‘late phase’.
Recent studies have begun to investigate the molecular basis for these two phases and their timing. For example, by eliminating certain early-phase proteins, it has been shown that they are not required for the recruitment of the late-phase machinery and vesicle budding, but instead control the spatial location of endocytic site assembly (Brach et al., 2014). Expression of the key early-phase protein Ede1 at sites typically devoid of CME results in the recruitment of downstream late-phase proteins (Brach et al., 2014). In addition, higher-order assembly of most of the late-phase proteins can be reconstituted in vitro by incubating microspheres that have been coated with late-phase Las17/WASp in yeast extracts. In these experiments, Las17/WASp-coated microspheres nucleated actin filaments and recruited most of the late-phase proteins, but interestingly, very few of the early-phase proteins (Michelot et al., 2010), supporting the conclusion that the early phase is separable from the late phase.
Because cargo molecules arrive after the early and early coat proteins, but before the late coat proteins (Toshima et al., 2006), it has been proposed that the transition from irregular to the regular lifetimes of later-arriving proteins might be controlled through a cargo checkpoint (Carroll et al., 2012), as has been proposed in mammalian cells (Loerke et al., 2009). This model is supported by data that show that the early-phase protein Ede1 has a much shorter lifetime in the bud of cells undergoing polarized growth, where it is thought that polarized secretion increases the amount of cargo in the bud, compared to that of the mother (Layton et al., 2011).
Key findings involving the late coat proteins Pan1, End3 and Sla1 have defined a site-initiation phase and an actin-polymerization phase (Sun et al., 2015). A pan1-null strain is inviable, but with the use of the auxin-inducible degron system, which acutely depletes proteins through proteasome-mediated degradation (Eng et al., 2014; Nishimura et al., 2009), Pan1 has been found to be required for the transition between early CME events and actin assembly (Bradford et al., 2015). Interestingly, when Pan1 and its interaction partner End3 are simultaneously depleted, the ‘early module’ and ‘early–mid coat module’ proteins are capable of colocalizing at the cell cortex; however, the WASp and myosin, scission and actin modules remain spatiotemporally uncoupled from these coat proteins (Sun et al., 2015). At the interface between these two functional phases, the cargo adapter Sla1 is recruited, which associates with Pan1 and End3. Although Sla1 is not the only CME cargo adapter, it is intriguing that it appears at the junction of these two phases, suggesting a possible role in coordinating cargo recruitment with actin assembly and vesicle formation. However, future studies are required to elucidate the mechanisms and functions underlying the coordination of these two phases.
Conclusions and perspectives
CME is a very regular and resilient process that has a very stereotypical assembly and disassembly timeline. Much work has been put into elucidating the timing and functions of the individual proteins, which we have now detailed in the accompanying timeline (see poster). However, to understand how CME functions and why it is so regular and predictable requires a deeper understanding of how the endocytic machinery is regulated and how each component interacts with the others. How is each step (site initiation, cargo capture, coat assembly, membrane bending, invagination, scission, etc.) regulated? What are the minimal components or key activities for each step? What are the biophysical properties of protein–protein or protein–lipid interactions, and how do they change as the site matures? Answering these questions will provide a basis for understanding how CME progresses. The findings are expected to apply to other organisms.
We thank Michelle Lu for helpful advice on poster design and Ross Pedersen for critical reading of the manuscript.
We also thank the National Institutes of Health (NIH) for funding [grant numbers R01 GM50399 and R01 GM42759]. Deposited in PMC for release after 12 months.
The authors declare no competing or financial interests.