Pioneering studies in the late 1970s revealed that clathrin-coated pits (CCPs) internalize different plasma membrane proteins and/or receptors and their ligands (cargos), including iron-containing particles (1), yolk proteins (2), immunoglobulin G (3), low-density lipoproteins [LDL; (4)], epidermal growth factor [EGF; (5)], α2 macroglobulin (6), insulin (7) and transferrin [Tf; (8–11)]. Some of these studies also revealed that a single CCP could contain several different cargos (12,13). The molecular basis of cargo capture by CCPs was established later, with the characterization of endocytic signals within the cytoplasmic domain of the internalized cargo [low-density lipoproteins receptor (LDL-R): (14); polymeric immunoglobulin receptor: (15); cation-independent mannose 6 phosphate receptor: (16); transferrin receptor (Tf-R): (17)]. These motifs were initially thought to interact directly with the assembly polypeptide 2 (AP-2) clathrin adaptor complex (18), which is involved in the assembly of clathrin triskelions at the plasma membrane (19,20). This simplistic view of AP-2 as the only CCP cargo-sorting protein was recently challenged by the discovery that certain receptors were internalized in CCPs even in cells depleted of functional AP-2 complexes (21), and by the characterization of novel cargo-specific adaptors (22). These findings suggest that there may be different, cargo-specific CCPs. This situation, if confirmed, raises many new questions: How do different sets of adaptors generate cargo-specific CCPs? How is the assembly of different CCPs regulated? How does cargo reach its specific CCP? In this review, we discuss the evidence for and against the existence of cargo-specific CCPs.
Because of the discovery of coated pits and vesicles more than 40 years ago and the identification of clathrin as a major component of the coat, it has been assumed that clathrin-coated pits (CCPs) are responsible for the uptake of most plasma membrane receptors undergoing internalization. The recent molecular characterization of clathrin-independent routes of endocytosis confirms that several alternative endocytic pathways operate at the plasma membrane of mammalian cells. This heterogeneous view of endocytosis has been expanded still further by recent studies, suggesting that different subpopulations of CCPs responsible for the internalization of specific sets of cargo may coexist. In the present review, we have discussed the experimental evidence in favor or against the existence of distinct parallel clathrin-dependent pathways at the plasma membrane.
The classical view: Constitutive versus ligand-induced clathrin-dependent endocytosis
The influence of ligand binding on receptor endocytosis has classically been used as a means of distinguishing between two different types of behavior in cargo internalization. In ‘constitutive endocytosis,’ all receptors are internalized, whether bound to their ligands or not. Receptors taken up in this way include nonsignaling receptors mediating the uptake of nutrients, as best exemplified by the LDL-R (23) and the Tf-R (24). In contrast, during ‘ligand-induced endocytosis,’ receptors are internalized only upon activation resulting from the binding of their specific ligand or agonist. Receptors taken up in this way include most of the known growth factor receptors, such as epidermal growth factor receptor (EGF-R) (145), although recent studies have shown that some signaling receptors may be internalized constitutively through CCPs independently of ligand binding (25,26). G-protein-coupled receptors (GPCR) comprise another large family of receptors in which endocytosis is initiated by agonist binding (reviewed in 27–29). A straightforward explanation for these different types of behavior is that endocytic signals operate differently in ligand-induced cargos, in which they may be masked and/or depend on posttranslational modifications. The internalization of GPCRs also requires additional adaptors [beta-arrestins (βarrs), see below] that are recruited in an agonist-dependent manner and connect activated receptors to the clathrin coat. Alternatively, the endocytosis of these two different populations of receptors may involve two different sets of CCPs.
The historical AP-2 complex: Assembly and sorting
The AP complexes were so called because of their intrinsic ability to stimulate the assembly of clathrin triskelions in vitro (146). They consist of four subunits, including two large ‘adaptins’ (α and β2 for AP-2) and two smaller subunits (μ2 and σ2 for AP-2), each involved in different functions (reviewed in 30,31). The AP-2 complex was predominantly localized at the plasma membrane (19), indicating that it could play a central role in the assembly of CCPs. This role was first suggested by studies in which neutralizing antibodies against α-adaptin were microinjected into cells (32) and was definitively confirmed by the recent observation that knocking down AP-2 expression by a small interfering RNA (siRNA) approach strongly decreases (by >90%) the number of CCPs at the plasma membrane (21,33). In classical models of CCP assembly, AP-2 is first recruited at the plasma membrane, where it nucleates the assembly of clathrin triskelions to form flat clathrin lattices or shallow CCPs (reviewed in 34,35). The clathrin lattice grows further by continually incorporating clathrin and AP-2 and then invaginates to form a sealed clathrin-coated vesicle (CCV), this final step requiring the GTPase activity of dynamin (Figure 1).
In addition to assembling CCPs, AP-2 also plays a key role in the sorting function of CCPs. Receptors are internalized through CCPs following the recognition of specific motifs located in their cytoplasmic domain. These ‘endocytosis motifs’ are necessary and sufficient for the clustering of the cargo in CCPs and its subsequent efficient internalization. Three major subtypes of endocytic motifs have been described. Tyrosine-based motifs, such as NPxY and Yxxφ, were first identified in the LDL-R and the Tf-R, respectively. Leucine-based motifs (LL, dileucine) are created by repeats of two hydrophobic amino acids (L, I, V, M) and are mostly found in the cytoplasmic tails of immune receptors. These two types of cargo are sorted by the AP-2 complex through the direct interaction of its μ2 subunit and β2 or α/σ2 hemi complexes with Yxxφ and dileucine motifs, respectively (reviewed in 36 and 37,38, respectively). As expected, AP-2 depletion strongly inhibits the internalization of cargos bearing these two types of signal (21,33,39). Interestingly, tyrosine and dileucine endocytic signals do not compete with each other for internalization (40,41) or for binding to purified AP-2 complexes (39), suggesting that they use different molecular devices for sorting in CCPs, despite both binding to AP-2. These results suggest that individual AP-2 complexes can independently and simultaneously sort several different subsets of cargo through interactions involving different subunits or subunit subdomains (see below and Figure 2).
New adaptors, New CCPs?
The use of siRNA to deplete endogenous AP-2 confirmed the crucial role of AP-2 in CCP assembly, but also revealed that some cargo could still be internalized in a clathrin-dependent process in the absence of AP-2 (21). These unexpected results, together with the observation that there were still a few CCPs present at the plasma membrane in these cells, raised the possibility that some CCPs can be formed in the absence of AP-2 and that these AP-2-independent CCPs are sufficient for sorting specific sets of cargo. This discovery also raised a new series of questions. How are these CCPs assembled and how are their cargos sorted? How different are they from AP-2-positive CCPs?
Over the past 10 years, an increasing number of new partners of AP-2 and/or clathrin have been identified and characterized (reviewed in 22,42,43). These novel AP-2-interacting proteins can be classified into two groups – accessory proteins involved in the CCP/CCV formation process itself (Eps15/Eps15R, epsins, CALM/AP180, amphiphysin, Hip1/Hip1R, SNX9, etc.) and alternative adaptors involved in the recognition and sorting of specific sets of cargo [autosomal recessive hypercholesterolemia (ARH), disabled-2 (Dab2), βarrs, Numb, Stonin, etc.]. Some, like the AP-2 complex itself, may be involved in both processes (see below). Alternative adaptors may mediate the sorting of receptors with endocytic signals that do not interact directly with AP-2. For example, the ARH and Dab2 proteins are responsible for internalizing receptors from the LDL-R family through interactions with NPxY motifs (see below). The Eps15/Eps15R and epsins are thought to mediate the recruitment of ubiquitinated cargos by interaction with their conserved [ubiquitin-interacting motif (UIM), see below]. In addition, βarrs interact with agonist-activated GPCRs and are required for the internalization of these receptors by CCPs. Finally, proteins such as Stonin 2, which is specifically involved in the AP-2 and clathrin-dependent internalization of synaptotagmin in neurons (44,45), may provide cell type specific sets of adaptors required for sorting of tissue-specific cargo.
One key question concerns whether these adaptors, also referred to as clathrin- and lipid-associated sorting proteins [CLASPs, (22)], are responsible for assembly of the functional CCPs resistant to AP-2 depletion. Some of these adaptors (e.g., Dab2 and epsins) have three properties in common with AP-2: (i) interaction with plasma membrane phospholipids, (ii) an ability to polymerize clathrin in vitro and (iii) the recognition of endocytic signals. Such a role would provide a simple mechanistic explanation for the cargo specificity of the AP-2-independent CCPs, as the sorting capacities of these CCPs would be directly linked to the cargo specificity of the alternative adaptors.
Constitutive and ligand-induced endocytosis appear to be mediated by morphologically identical CCPs at the plasma membrane. Although several studies have indicated that markers of both pathways (EGF versus LDL or Tf) can be found within the same CCPs (12,13,46), the coexistence of the two different types of endocytic behavior was seen as strong evidence for distinct populations of CCPs mediating different types of internalization. By extension, the possibility that receptor subtypes could be sorted by specfic CCPs was then applied to both constitutive and ligand-induced endocytosis.
Cargos internalized by constitutive endocytosis
Constitutively, internalized cargos were thought to participate in CCP assembly by stabilizing the docking of AP-2 complexes at the plasma membrane. This model is supported by recent observations showing that assembling CCPs are stabilized by the incorporation of cargo (47). In addition, studies in an in vitro system have clearly indicated that the presence of peptides mimicking cargo cytoplasmic domains increases the binding of AP-2 to liposomes (39). These data suggest that cargos for constitutive endocytosis may somehow participate in CCP assembly, thereby influencing the final content of the CCP they have helped to stabilize or assemble. The recent discovery of adaptors or regulators specific for some of these cargos (Tf and LDL) also provides support for this model.
The prototype: Tf and its receptor
The Tf-R is the most widely used model for studies of ‘general’ clathrin-dependent internalization, and it is currently accepted that Tf-R, which interacts with the μ2 subunit of AP-2 through its YTRF motif, may be present in every single CCP at the plasma membrane, as suggested by the distribution of these two proteins at the plasma membrane by electron (see above) and fluorescence microscopy [(48–50) and our unpublished observations]. As expected, the depletion of AP-2 or the clathrin heavy chain strongly inhibits the Tf internalization (21,33,51,52).
Despite the body of data indicating that Tf-R may be used as a general marker of the clathrin-dependent pathway, some studies have indicated that the endocytosis of this receptor may be specifically regulated. The Tf-R overexpression increases the number of CCPs at the plasma membrane (53), suggesting that this cargo may regulate CCP assembly. It was later shown that Tf-R overexpression promotes the assembly of flat clathrin lattices (54). The recent description of a specific regulator of Tf-R endocytosis, transferrin transport protein [TTP, (55)] has provided evidence that constitutive and ligand-induced endocytosis can be distinguished at the molecular level. The TPP has been shown to interact with both the cytoplasmic domain of Tf-R and dynamin. The TTP controls the GTPase activity of dynamin, and this negative control is lost upon binding to the cytoplasmic domains of Tf-R, suggesting that Tf-R may control the final budding event of the CCPs to which it has been recruited. Furthermore, TTP is found in Tf-R-containing CCPs but not in EGF-R-containing CCPs, implying that TTP may define a subset of TF-R-positive CCPs. The EGF-induced phosphorylation of either dynamin or TTP, preventing their mutual interaction, has been identified as a possible mechanism for the exclusion of TTP, and thus of Tf-R, from EGF-R-positive CCPs. It remains unclear whether the TTP-containing CCPs sort or exclude other cargos in the constitutive endocytic pathway. These data, together with the finding that other cargos may be internalized in a clathrin-dependent manner even in the absence of AP-2 (21), define Tf as the marker of AP-2 and clathrin-dependent endocytosis.
LDL and its receptor
The LDL-R was the first receptor to provide insight into the molecular mechanisms behind the sorting of competent cargos by CCPs (reviewed in 56). Indeed, pioneering studies by Brown and Goldstein showed that the NPxY tyrosine-based signal in the cytoplasmic tail of the LDL-R, which is mutated in familial hypercholesterolemia patients, was required for efficient internalization through CCPs (14). However, the mechanisms by which this type of signal is interpreted by the endocytic machinery remained poorly understood until very recently. The NPxY motif of the LDL-R differs from classical YxxF-based endocytic signals as it binds the μ2 subunit of AP-2 only weakly in various systems (57,58). This motif is sorted by two specific adaptors, ARH and Dab2, which were thought to act together to mediate internalization through CCPs upon interaction with both AP-2 and clathrin (59–63). It therefore came as a surprise to find that internalization of a chimera bearing the cytoplasmic domain of the LDL-R was strongly inhibited by downregulation of endogenous clathrin heavy chain expression but not by downregulation of endogenous AP-2 expression (21). The requirement for clathrin but not for AP-2 for LDL-R internalization suggested that LDL-R was internalized through a subpopulation of CCPs that did not require AP-2 for assembly. This raised questions about whether these CCPs were assembled by the LDL-R-specific adaptors Dab2 and/or ARH.
Two recent studies investigated the respective roles of AP-2, ARH and Dab2 in the clathrin-dependent endocytosis of LDL/LDL-R (50,64). Both studies confirmed that AP-2 is dispensable for the clathrin-dependent internalization of LDL-R. The ARH-mediated internalization of LDL was dependent on AP-2, whereas Dab2 mediated the clathrin-dependent internalization of LDL-R in the absence of AP-2. These results suggest that Dab2 can assemble a subset of AP-2-independent CCPs at the plasma membrane, consistent with its ability to interact with membrane phospholipids and to polymerize clathrin in vitro (63). However, endogenous Dab2, like ARH, was found to colocalize strongly with AP-2 in various cell types (50,60), suggesting that, under normal conditions (i.e., in the presence of AP-2), Dab2 acts as a specific LDL-R adaptor targeted to AP-2-containing CCPs. In the absence of AP-2, Dab2 is then likely involved in the assembly of and/or is recruited to the 5–10% CCPs remaining in AP-2-depleted cells (21,33). It remains unclear how these CCPs support a normal rate of LDL uptake. It may be that alternative adaptors, which are normally evenly distributed among CCPs (50), become more concentrated in the remaining CCPs, thereby increasing the concentration of the cargo in fewer structures (Figure 2). However, although these issues are crucial to our understanding of events in cells lacking AP-2, they remain to be directly investigated.
In yeast, cargo ubiquitination is the main signal for efficient endocytosis and sorting in endosomes. Cargo ubiquitination has often been reported to play an important role in targeting internalized cargos for degradation in late endosomes in mammalian cells, but it remains unclear whether ubiquitination could act as a general signal for internalization (reviewed in 65,66). Nevertheless, the ubiquitin peptide has clearly been shown to support constitutive clathrin-dependent endocytosis when fused to reporter molecules (67–70). The proteins responsible for the active sorting of ubiquitin-bearing cargos were recently identified as Eps15/Eps15R and epsins. These proteins, initially identified as CCP-associated proteins (71–75), were shown to contain a highly conserved UIM for direct interaction with ubiquitin (76) and the ubiquitin-related protein ubiquilin/hplic (77). These proteins may therefore be considered to be alternative adaptors that can efficiently sort ubiquitinated cargos into CCPs, as described for their orthologs in yeast (78). The internalization of reporter molecules with ubiquitin fused to their cytoplasmic domains was recently shown to be inhibited by the depletion of clathrin, epsin or Eps15/Eps15R but not by the depletion of AP-2 (69). Interestingly, the depletion of epsin and clathrin, but not of AP-2, has also been reported to inhibit the ubiquitin-dependent internalization of major histocompatibility complex class I molecules in cells expressing the K3 ubiquitin ligase of the Kaposi's sarcoma-associated herpesvirus (79), suggesting that an AP-2-independent, clathrin- and epsin-dependent pathway may be involved in the internalization of physiological cargos. Indeed, the targeting of epsin functions alone or together with those of Eps15/Eps15R, by siRNA techniques or the overexpression of mutant forms, has been reported to affect the internalization of ubiquitinated receptors, including the dopamine receptor (80) and the epithelial sodium channel (81).
These results suggest, as for LDL-R, that ubiquitinated cargos may be sorted independently of AP-2 by CCPs assembled with ubiquitin-specific alternative adaptors. The Eps15/Eps15R and epsin may play a role similar to that of Dab2 for LDL-R. The targeting of Eps15 to CCPs appears to be AP-2 dependent (33,70,82), whereas epsin was found to colocalize with the clathrin-positive CCPs observed in AP-2-depleted cells (21,33,70). Epsin has also been shown to be sufficient to drive the in vitro assembly of clathrin on phospholipid monolayers (83), suggesting that epsin may replace AP-2 in CCP assembly, and that the resulting subsets of CCPs may be functional for the sorting of ubiquitinated cargos (Figure 2).
Ligand-induced specific CCPs
By analogy with the recruitment of signaling molecules to activated growth factor receptors (Figure 3), it has been suggested that the activation of signaling receptors could also recruit the molecular machinery for de novo CCP assembly for their own internalization. The de novo formation model was widely seen as a valid model reflecting the ligand-induced internalization of EGF-Rs and GPCRs. However, using another receptor model for ligand-induced endocytosis (high-affinity immunoglobulin E receptor), Santini et al. provided evidence that CCPs are not assembled at the site of receptor activation and challenged the de novo assembly model for signaling receptors (84). An alternative model would be that activated receptors are targeted to CCPs either already formed or in the process of assembly before receptor activation (preexisting CCPs). These two models, which remain a matter of debate, involve major differences in CCP behavior. De novo CCP assembly implies that receptor activation regulates the CCP assembly machinery and that newly formed CCPs are probably restricted to the stimulated receptor. The preexisting model implies that receptor activation controls cargo recruitment to CCPs but not CCP assembly, and suggests that activated receptors share CCPs with cargos taken up by constitutive endocytosis.
GPCRs, targeting to preexisting CCPs?
The de novo assembly model was put forward following studies of the endocytosis of GPCRs. In this particular case, GPCR internalization requires additional molecules – βarrs (βarr1 and βarr2) – that were initially identified as proteins required for desensitization of the β-adrenergic receptor. βarrs bind to GPCRs upon agonist activation, and interact directly with both clathrin and AP-2 (85–87). These findings raised the possibility that GPCR/βarr complexes could drive the assembly of their own CCPs in response to receptor activation (Figure 3). However, live-cell imaging studies combining different GPCRs (thyrotropin-releasing hormone, β2-adrenergic and dopamine receptors), with either βarr1 or βarr2, and various markers for CCPs (Eps15 or clathrin light chain) clearly showed that GPCR/βarr complexes were recruited to CCPs assembled before agonist stimulation (88–90).
It remains possible that the preexisting CCPs to which activated GPCRs are recruited differ from other CCPs. If this were the case, only some of the preexisting CCPs would be competent for the recruitment of activated GPCRs. Initial studies in live and fixed cells overexpressing both GPCRs and βarrs showed that βarr/GPCR complexes were recruited to almost all of the CCPs present at the cell surface [(88–90) and unpublished observations] and that internalizing GPCR were found into structures that were positive for Tf/Tf-R (89,91). Interestingly, Puthenveedu and Von Zastrow, using a similar approach, found that the proportion of CCPs containing activated GPCR decreased to 50% if only GPCR was overexpressed (90), consistent with the results of another recent study (92). These results suggest that GPCR/βarr complexes are targeted to specific subsets of preexisting CCPs. However, the observation that the proportion of CCPs containing activated GPCR increases with βarr levels suggests that GPCR/βarr complexes can be recruited to all preexisting CCPs, the extent of this recruitment being limited by the amount of cytoplasmic βarr available. This is consistent with the requirement for binding to both AP-2 and clathrin for the targeting of βarrs to CCPs (93,94), and with the fact that AP-2 and clathrin binding only could not provide the information for targeting to a specific subset of CCPs. As expected, mutated forms of βarrs, which mimic activation by agonist, are found in most if not all CCPs at steady state (94), indicating that βarrs are unlikely to represent a targeting device for specific CCPs.
EGF-R, the de novo CCP cargo prototype?
Most of the support in favor of the de novo model of CCP formation comes from studies on endocytosis of the EGF-R. Unoccupied EGF-Rs are not efficiently internalized, whereas EGF-Rs are rapidly taken up into CCPs after activation. Indeed, after much debate (95,96), it was finally established that the tyrosine kinase activity of the EGF-R is required for the recruitment of EGF-R/EGF complexes into CCPs (97). Furthermore, the EGF-R is one of the very few receptors that can be coimmunoprecipitated with the AP-2 complex (98). The occurrence of this interaction upon EGF activation has been seen as strong evidence for EGF-driven de novo assembly of EGF-specific CCPs through the specific recruitment of AP-2 to activated EGF-R (Figure 3).
This hypothesis was challenged by the observation that mutated EGF-R unable to bind to AP-2 was nonetheless internalized (99). Similarly, cells depleted of AP-2 or expressing a mutated form of the μ2 subunit unable to bind activated EGF-R have been shown to internalize EGF (21,100), suggesting that CCPs devoid of functional AP-2 remain competent for EGF-R endocytosis. The apparent dependence of EGF-R internalization on clathrin under the same conditions (21,100), suggests that activated EGF-R may be internalized through AP-2-independent CCPs, like the LDL-R. However, it has also been reported that EGF internalization may be affected by AP-2 depletion (52,101). These apparent discrepancies may be because of differences in the experimental conditions used in these studies. Indeed, factors such as cell type, EGF concentration and stimulation conditions may determine the endocytic pathway taken by activated EGF-R (102–106). Thus, clathrin- and AP-2-dependent internalization of the EGF-R occurs when the receptor is directly stimulated by a low dose of EGF (∼1–3 ng/mL), at 37°C, in cells not starved of serum (52). Under these conditions, EGF internalization is inhibited or delayed by the depletion of endogenous AP-2, clathrin heavy chain or UIM-containing proteins, including Eps15/Eps15R and epsins (52,101,105).
Other reports have suggested that EGF treatment induces the assembly of EGF-specific CCPs. An initial study by Brodsky et al. reported that at high doses of EGF (250 ng/mL), with prebinding at 4°C, in serum-starved cells, rapid tyrosine phosphorylation of the clathrin heavy chain occurred in a c-src-dependent manner, resulting in the massive recruitment of clathrin to the plasma membrane, and correlated with an increase in EGF uptake (107). These observations suggest that, under these conditions, activated EGF-R can drive the assembly of new CCPs. This hypothesis was further confirmed by an electron microscopy analysis, which showed that the number of CCPs doubled in cells stimulated by EGF at 4°C. These CCPs contained EGF-R, AP-2 and clathrin, together with EGF-induced signaling molecules (108). It has also recently been shown that EGF-R activation increases the activity of the phosphatidylinositol kinase type I, which produces PI(4,5)P2, the phosphoinositide specifically involved in AP-2 recruitment at the plasma membrane (109). These results suggest an alternative mechanism through which EGF treatment may activate the molecular machinery required for the de novo assembly of AP-2-containing CCPs. The EGF-driven assembly of CCPs has also been observed in cells depleted of AP-2 (101). This clearly indicates again that CCP can be formed in the absence of AP-2, however, it does not demonstrate that this process occurs in ordinary conditions.
Thus, the incubation of cells with EGF at 4 °C can induce the assembly of new CCPs, in a process that may be independent of the AP-2 complex. In addition, EGF-induced CCPs appear to be specific to the EGF-R, or at least they contain fewer Tf-R than the ‘classical’ CCPs found in unstimulated cells (101,108) and contain markers of rafts (GM1) normally excluded from Tf-R-containing CCPs (108), suggesting that they are different from the CCPs involved in constitutive endocytosis. These recent studies are consistent with previous findings that the molecular machineries involved in the recruitment of Tf and EGF receptors to CCPs differ (46). They are also consistent with the convincing demonstrations that EGF-R does not compete with Tf-R for endocytosis, regardless of the EGF concentration used or the number of EGF-R expressed at the cell surface (103,104,110). Similarly, dileucine and tyrosine-based signals do not compete with each other for internalization. The lack of competition between Tf-R and EGF-R for endocytosis implies only that the two receptors do not compete for the same sorting device – the μ2 subunit of AP-2. Finally, the absence of a positive effect of EGF stimulation on Tf endocytosis (103,104), despite the induction of new CCP formation, suggests that the newly formed CCPs cannot sort Tf-R. This is consistent with electron microscopy data (see above) and the absence of TTP in EGF-R-positive CCPs (55).
As AP-2 can be dispensable for EGF-R clathrin-dependent endocytosis, alternative adaptors or a specific molecular machinery may exist for the recruitment of activated EGF-R, as described above for LDL-R or ubiquitin. Several proteins are good candidates for this role. Perforated cell assays have shown that a substrate of the EGF-R tyrosine kinase is required for the efficient recruitment of activated EGF-R, but not of Tf-R, into CCPs (97). This substrate may be Eps15, as it was later shown that the tyrosine phosphorylation of Eps15 was required for efficient targeting of EGF-R, but not of Tf-R to CCPs (111). Eps15 does bind AP-2 (112), but is unlikely to substitute for AP-2 in the assembly of CCPs, as it cannot interact with clathrin (113), and its incorporation into CCPs is dependent on AP-2 (33,82). Interestingly, downregulation of CALM, the ubiquitous form of neuronal AP180, has been shown to inhibit the endocytosis of EGF-R, but not that of Tf (52), suggesting that this protein plays a specific role in EGF-R endocytosis. CALM is found in CCPs at steady state (114), and AP180 has been reported to be sufficient for clathrin assembly on a lipid monolayer (115). CALM is therefore a good candidate for involvement in the assembly of EGF-R-specific CCPs. However, further studies are required to understand the precise role of this molecule, as overexpression of CALM mutants has blocked Tf-R endocytosis (114).
The recruitment of EGF-R to CCPs also appears to require proteins previously identified as involved in signal transduction. For example, Grb2 has been shown to be required for the endocytosis of EGF (116) during the recruitment of activated EGF-R to CCPs (52,101,117). The precise mechanism of action of Grb2 remains unclear, but may involve the interaction of this molecule with the ubiquitin ligase Cbl (118,119). It remains unclear how this complex becomes linked to a clathrin assembly protein responsible for the final assembly of CCPs. It was initially suggested that the ubiquitination of EGF-R might be involved. However, neither the ubiquitin ligase activity of Cbl nor the ubiquitination of the EGF-R itself are required for CCP clathrin-dependent endocytosis of the EGF-R (118,120). Identification of the final adaptor for EFG-R would provide real insight into this complex internalization process.
Other signaling receptors possibly involved in de novo CCP assembly
A few other receptors can, like EGF-R, induce the assembly of their own endocytic structures. One example is the nerve growth factor receptor (NGF-R). Early studies indicated that the treatment of NGF-starved cells with NGF or EGF increased the total number of CCPs present at the plasma membrane (121,122). In this case, the activation of the NGF-R results in an increase in AP-2 and clathrin levels at the plasma membrane and an increase in clathrin heavy chain phosphorylation (123), as observed for EGF-R (see above). However, in contrast to what has been observed for EGF, NGF treatment also stimulates Tf internalization (123), suggesting that the NGF-induced CCPs do not differ from the CCPs involved in constitutive endocytosis.
An increase in clathrin recruitment at the plasma membrane and/or in tyrosine phosphorylation of the clathrin heavy chain has also been described for the insulin receptor (124,125) and for the B-cell and T-cell receptors [BCR and TCR, respectively; (126,127)]. Ligation of the BCR and TCR results in the tyrosine phosphorylation of clathrin and the endocytosis of these receptors through a clathrin-dependent internalization pathway. The BCR is internalized only if phosphorylated clathrin is associated with lipid rafts. The Tf-R was not found associated with these lipid rafts, suggesting that BCR endocytosis occurs through a different set of CCPs, as suggested for EGF-induced CCPs (108).
Thus, although it is clear that EGF treatment under specific conditions can induce the formation of CCPs that are specifically involved in the internalization of EGF-R, or at least not involved in the internalization of Tf, the existence of similar mechanisms for other receptors remains possible but unproven. Direct morphological studies and live-cell imaging would be useful and might make it possible to link clathrin modifications with changes in the CCP assembly process at the plasma membrane.
CCPs and pathogens
A number of viral pathogens enter cells by receptor-mediated endocytosis, targeting membrane receptors internalized by clathrin-dependent or clathrin-independent endocytosis (reviewed in 128). Two recent studies have investigated the dynamics of entry of the influenza virus (129) and reovirus (47) with respect to clathrin. In both studies, the virus was found to bind initially to the plasma membrane, on which it remained static. Clathrin staining was observed a few minutes later, at sites of virus attachment, followed by the endocytosis of the viral particles, as shown by the rapid disappearance of clathrin staining and rapid lateral movements of the internalized particles. In both cases, these events are consistent with the possibility that viruses may induce the assembly of their own clathrin-coated structures for internalization. These structures probably correspond to CCPs, as an early morphological study showed that influenza virus particles were present in CCPs and nascent vesicles (130).
A recent study unexpectedly showed that internalization of the bacterium Listeria monocytogenes was affected by depletion of the heavy chain of clathrin (131). Interestingly, an siRNA screen revealed that AP-2 was not involved, whereas dynamin, Eps15 and a number of signaling proteins, including Grb2, the met receptor and Cbl, were required for Listeria entry. These results suggest that Listeria may use the met receptor to induce the assembly of clathrin-coated structures similar to those induced by EGF-R stimulation (see above). However, bacteria are larger than classical CCPs, and electron microscopy studies are required to investigate this issue.
Electron microscopy has definitively documented an increase in the number of CCPs present at the cell surface in response to receptor stimulation (EGF or NGF) and has led to the identification of components of the endocytic machinery and to insight into the nature of the cargo present in newly formed CCPs. However, in fixed cells, it is impossible to determine whether the increase in the number of CCPs observed at the plasma membrane results from the stimulation of assembly or changes in the kinetics of invagination and/or budding. For instance, the accumulation of CCPs at the plasma membrane was recently observed following the transient chemical inhibition of dynamin activity (132).
Recent technical developments in cell imaging, making it possible to visualize the dynamics of receptors and clathrin structures in real time, should, as it has for GPCRs, make it possible to determine unambiguously whether the EGF-R stimulates the formation of its own CCPs or is recruited to preexisting CCPs. Real-time imaging of the sorting of cargos undergoing constitutive endocytosis and differing in size (LDL, Tf and rheovirus) has indicated that these cargos are recruited to newly formed CCPs (47). This study also revealed that the time required to assemble a new CCP was proportional to the size of the cargo, leading to the provocative suggestion that CCPs may begin to assemble in a random, constitutive manner but collapse in the absence of cargo. The EGF uptake was not assessed, but the results of this study suggest that the formation of new CCPs is constitutive rather than induced. Fluorescence-based techniques provide powerful tools but also have pitfalls making it difficult to interpret the results obtained. The need to overproduce fluorescent-protein-tagged versions of the proteins of interest may perturb the whole process of signal transduction and endocytosis, particularly for signaling receptors. In addition, receptors or CCP-associated proteins may display diffuse membrane staining, increasing the local background level of fluorescence over the level of fluorescence of bona fide CCP proteins, such as clathrin or AP-2. For example, the distribution of the HIV Nef protein has been analyzed by total internal reflection fluorescence microscopy (133). Nef was detected in only 60% of the clathrin-positive structures at the plasma membrane. This could have been interpreted as restricted recruitment of Nef in a subpopulation of CCPs. In fact, this result reflected the technical difficulty of detecting Nef in dimmer CCPs because Nef-associated fluorescence was equivalent to local membrane background in those CCPs. A similar problem was encountered in studies of the distribution of epsin with respect to that of clathrin (134).
The mechanistic point of view
Several studies have suggested that some cargos can drive the assembly of specific CCPs. However, this possibility is not supported by the observations that most known alternative adaptors do bind AP-2 and/or clathrin, and that AP-2 binding appears to be involved in their targeting to CCPs [βarr, Dab2, ARH: (60,87,135)]. Therefore, in the absence of identified cargo adaptors truly independent of AP-2, it is difficult to imagine how CCPs with distinct sorting capacities could be generated with the targeting of cargo-specific adaptors mediated through a common molecular device – AP-2 binding.
Indeed, immunofluorescence and electron microsocopy studies have shown that at steady state, alternative adaptors, such as Eps15, epsin, ARH and Dab2, are detected in almost all CCPs (50,71,75). This implies that every given CCP at the plasma membrane contains molecular devices for the sorting of different subtypes of cargo. All CCPs contain AP-2, which can itself actively and simultaneously sort both YxxF and LL signals through its μ2 and α/σ2 or β2 subunits. In addition, AP-2 interacts with epsin, Eps15 and Dab2 through the C-terminal domain of α-adaptin, resulting in the selection of ubiquitinated cargos and cargos bearing the NPxY signal (LDL-R). Finally, AP-2 interacts with both βarr and ARH through the C-terminal domain of β2-adaptin, resulting in the indirect sorting of activated GPCRs and cargos bearing NPxY signals (Figure 2). Thus, Grb2 and CALM/AP180 may be good candidates for mediating the recruitment of EGF-Rs to specific CCPs, but their precise mechanism of action requires more detailed definition.
The final content of a given CCP therefore probably depends on the level of receptor expression at the plasma membrane for constitutive endocytosis, on the level of both receptor expression and agonist concentration for ligand-induced endocytosis, and on the level of expression of each alternative adaptor required to recruit specific sets of cargos to CCPs. According to this model, a given cargo could be excluded from CCPs because of simple competition for alternative adaptors to AP-2 and/or clathrin. Indeed, ARH and βarr bind through a FxxFxxxR motif similar to that in the equivalent subdomain of β2-adaptin, with similar affinities (136,137). It is therefore tempting to speculate that, upon GPCR stimulation, βarr competes with ARH for β2-adaptin binding and then exclude LDL-R from GPCR-containing CCPs. This possibility remains to be tested. However, Dab2 can mediate clathrin-dependent endocytosis of LDL-R independently of AP-2, suggesting that this molecular competition for β2-adaptin probably occurs only in cells that express ARH but not Dab2, as shown in hepatocytes (50,64). Another possibility was provided by the discovery of TTP. In this case, activation by EGF would exclude Tf-R from CCPs induced by EGF (55). Future studies are required to determine whether there are other CCP ‘excluders’ for other cargos, such as GPCRs.
Signaling receptors, such as the EGF-R, may regulate the interaction of alternative adaptors with the AP-2 complex and/or the sorting capacities of these alternative adaptors. The EGF activation induces the tyrosine phosphorylation of both Eps15 (111) and β2-adaptin (138) and the monoubiquitination of Eps15 and epsin (76,139). The effects of these modifications are unclear, but they may regulate activity. In the case of epsin and Eps15, these modifications may influence sorting properties for ubiquitinated cargos.
By analogy with the CCVs formed in the trans Golgi network, which differ from those originating at the plasma membrane in terms of the subtype of AP complex – AP-1 versus AP-2 – it seems likely that different populations of CCPs at the plasma membrane are defined by different sets of alternative adaptors (epsin, Dab2/ARH, CALM/AP180). Although this appears possible in cells depleted of endogenous AP-2, there is no evidence for the existence of CCPs containing epsin, CALM or Dab2 but no AP-2 in ordinary conditions. The only clear detailed example of the specific assembly of CCPs by EGF is restricted to certain experimental conditions. This raises questions as to whether EGF is a unique example of a growth factor able to induce the assembly of CCPs or whether this induction is a more general feature of signaling receptors. The GPCRs, the largest family of signaling receptors, do not seem to induce CCP assembly (see above). There is evidence to suggest that activation of other receptors can induce the assembly of CCPs. However, a system in which each signaling receptor is able to induce its own structure for internalization is unrealistic, unless the different receptors share common machineries or mechanisms.
The strongest evidence for the existence of different CCPs is provided by real-time imaging experiments in which the dynamics of clathrin and other CCP-associated proteins have been visualized in living cells. These studies have revealed a high degree of diversity among the CCPs of a given cell (140). Most CCPs were visualized as highly dynamic spots forming vesicles within minutes, whereas a subset of CCPs (up to 25%) appeared to remain static for 10–30 minutes (49,134,141), and to be able to undergo multiple rounds of internalization (49,141). The dynamic population corresponds to the classical view of CCV formation, whereas this more static population was not predicted. Such behavior has been reported in various cell lines, including primary cells (141), but has not been observed or analyzed in others (47). These data highlight an unexpectedly high degree of complexity and suggest differences in the mechanisms of CCP/CCV formation (reviewed in 142,143), possibly related to cell-type specificities in endocytosis. How and why such differences are generated remains a matter of speculation. These differences are highly consistent with the characteristics of signaling receptors. Indeed, receptor sorting and signal transduction are known to be tightly coordinated for a number of signaling receptors. Thus, the extraordinary complexity and promiscuity of signal transduction may result partly from the generation of different CCPs and features specific to particular cell types. Different CCPs may also transport cargos to different endosomes for different cellular functions (144). It should be possible to test these elusive hypotheses when specific cargo-dependent regulators of recruitment to CCPs are identified either as adaptors or excluders.
The work in the laboratories of the authors was supported by grants from the ‘Association pour la Recherche sur le Cancer’ [n° 36–91 (A. B.) and n° 31–43 (C. L.)] and from the ‘Agence Nationale de la Recherche’ (C. L.). The authors wish to thank Mark G. Scott, Joshua Z. Rappoport, Ludger Johannes and Inger Helene Madshus for critical reading of the article.