Ever since the discovery that hormone and growth factor receptors are internalized following ligand binding, scientists have asked ‘why?’ Historically, receptors were thought to reside exclusively at the cell surface where they were in a position to sample the extracellular environment. The notion of an activated receptor being rapidly transferred to an intracellular compartment was intriguing and seemed full of interesting possibilities with respect to receptor function. The ensuing decades have witnessed many efforts to unravel the role of endocytosis in receptor signaling, but many issues remain unresolved. Certainly, early hypotheses that endocytosis delivered activated receptors to the nucleus have found little support. However, our deepening understanding of the complexities of signaling networks has provided many new and interesting hypotheses. In addition, our improved understanding of endocytosis in general has provided a valuable context for understanding regulated receptor trafficking. In this review, we will cover some of the most recent findings regarding the trafficking of receptor tyrosine kinases (RTKs) and the role this plays in regulating signal transduction.
Activated receptor tyrosine kinase (RTK) receptors are rapidly internalized and eventually delivered to the lysosomes. Although ligand-induced endocytosis was originally thought to be a mechanism of receptor inactivation, many studies suggest that receptors remain active within endosomes. This review discusses the role that internalized signaling complexes may play in different RTK systems including recent data on how ubiquitination may regulate this process. In general, it appears that some receptor systems have evolved to enhance endosomal signaling, as is the case for TrkA and NGF. In contrast, the insulin receptor system appears to limit the extent of endosomal signaling. The EGFR system is the intermediate example. In this case, some signals are specifically generated from the cell surface while others appear to be generated from within endosomes. This may act as a mechanism to produce ligand-specific signals. Thus, trafficking could play diverse roles in receptor signaling, depending on the specific cell and tissue type.
Possible Roles of Receptor Internalization
Initially, it was observed that, following ligand addition, the epidermal growth factor receptor (EGFR) underwent rapid endocytosis . This observation has been extended to virtually all signaling receptors. The most obvious function that endocytosis can serve is to remove activated ligand–receptor complexes from the cell surface to allow cells to receive additional signals. Ligands must be consumed as an intrinsic part of signal transmission. Otherwise, cells could not discriminate new signals from old ones. Endocytosis could obviously fulfil this role by targeting activated receptor–ligand complexes to the lysosomes, but it is unclear whether endocytosis plays additional roles. Trafficking of activated RTKs is a complex and highly regulated process, consistent with more than a single role in signaling.
There are two types of functions that receptor trafficking could have in regulating receptor signaling. It could control the magnitude of the response or it could control the specificity of the response. It has been known for many years that the majority of activated EGFRs are located in endosomes . If intracellular receptors are capable of triggering the same signaling cascades as surface-localized receptors, then regulating the size of the endosomal pool would modulate the magnitude of the response. Alternatively, if the signals generated by receptors in endosomes are distinct from those generated at the cell surface, then regulated trafficking could alter the specificity of the receptor response. Although seemingly simple, discriminating between these two possibilities has proven difficult. Resolving the ‘quantity versus quality’ issue is complicated by both conceptual and experimental problems. Conceptually, we do not fully understand what dictates signal specificity. If two different RTKs phosphorylate a distinct set of substrates, we expect them to generate a different set of biological responses. But what happens when you change the magnitude or length of signaling? For example, it has been proposed that a short duration of signaling in PC12 cells will promote proliferation whereas prolonged signaling leads to differentiation . In the case of the EGFR, preventing EGFR down-regulation facilitates cell transformation . Thus, changing signaling kinetics and magnitude can change cellular responses. However, most of these studies use experimental manipulations that could potentially alter the pattern of substrate phosphorylation or alter endocytic trafficking. Receptor trafficking takes time and is both a specific and a saturable biochemical process [5,6]. Thus, changing the length and magnitude of ligand exposure will change the distribution of receptors between the cell surface and the endosomal compartments, which, in turn, could alter the pattern of substrate phosphorylation.
For both theoretical and practical reasons, it appears most likely that endosomal trafficking regulates signaling specificity by controlling the pattern of substrate modification. Signal duration and frequency can indeed regulate gene expression at the transcriptional level, but the only physiological systems where this has been documented involve regular, pulsatile patterns of hormone release, such as growth hormone production in the rat . There is little evidence that growth factor levels are rapidly modulated in situ and there are no documented examples where alterations of growth factor production rates in vivo give rise to qualitatively different biological outcomes. In contrast, there is increasing evidence that substrates of RTKs have distinct distributions within cells, even differing between the apical and the basolateral surfaces of polarized epithelial cells . These differences in substrate location could affect cellular responses.
Obtaining experimental evidence of a role for endosome-specific substrate phosphorylation in specific biological responses has proven elusive. This is primarily because one cannot get activated receptors inside a cell without first passing through the cell surface. Thus, one can always argue that any given response was initiated at the cell surface. Biochemically, it has proved extremely difficult to rapidly separate the cell surface and endosomal membranes and, thus, to identify the site of signal generation. Furthermore, our poor understanding of the relationship between the phosphorylation of a given RTK substrate and subsequent biological responses has made it difficult to design appropriate experiments.
Fortunately, some recent approaches have made the answers to these questions more accessible. One such approach has been the development of dominant-negative and temperature-sensitive mutants of dynamin that allow coated pit endocytosis to be selectively inhibited . If blocking endocytosis has a significant effect on signaling, one can infer that endocytic trafficking is important. These and other approaches have been used to examine endosomal signaling from the EGFR , the insulin receptor (IR) , and the TrkA receptor . These recent studies provide strong evidence that receptor trafficking is an important mechanism for regulating the specificity of RTK-generated responses. However, the significance of endosomal signaling appears to be very dependent on both the receptor and the cell type. This is likely because different receptor systems are functionally adapted to their specific cellular contexts. Endosomal signaling may be very important in some receptor systems and relatively unimportant in others. This review considers three receptor systems in which endosomal signaling has been recently investigated.
The EGFR System
Regulated trafficking of the EGFR has been extensively investigated over several decades and probably represents the best understood receptor trafficking system (see Fig. 1). Following ligand binding, the EGFR is rapidly internalized by a mechanism that requires intrinsic receptor kinase activity and specific motifs in the carboxy terminus domain of the receptor. The mechanistic basis for the kinase requirement for EGFR internalization is unknown. Substrates such as Eps15 have been proposed to be EGFR-specific recruiting factors , but Eps15 lacks the appropriate specificity. For example, the expression of a dominant-negative form of Eps15 inhibits internalization of both EGFRs and transferrin receptors due to disruption of coated pit formation . However, EGFRs and transferrin receptors have been shown to use different recruiting factors .
Following internalization, the EGFR is sorted into early endosomes and late endosomes. This initial sorting step appears to be mediated by di-leucine motifs in the juxtamembrane region of the receptor . Interestingly, the human adenovirus E3-13.7 protein induces down-regulation of the EGFR by binding to this region in early endosomes and facilitating receptor transport to late endosomes . Thus, the viral protein appears to mimic the action of the early endosome sorting protein, the identity of which is currently unknown. Once in the late endosomes, the EGFR is either targeted to lysosomes or recycles back to the cell surface . The lysosome-targeting sequences of the EGFR also include a di-leucine motif, but are located distal to the kinase domain . In contrast to the requirement of kinase activity for occupancy-induced recruitment into coated pits, regulated endosomal sorting appears to independent of receptor kinase activity [19,20]. Instead, occupancy-induced conformational changes in the receptor appear to be sufficient. Receptor kinase activity is involved to the extent that it directly, or indirectly, influences receptor conformation. Thus, there are three distinct steps in regulated EGFR trafficking: endocytosis, early endosome sorting, and lysosomal targeting. Each of these steps has distinct structural and biochemical requirements. These sorting steps appear to be common to all RTKs, although their degree of regulation is receptor-specific. For example, platelet-derived growth factor (PDGF) receptors are efficiently targeted to lysosomes even in the absence of the cognate ligand or receptor kinase activity .
The EGFR family of RTKs includes the EGFR as well as erbB-2, erbB-3, and erbB-4 . The human forms are also known as HER1, HER2, HER3, and HER4. Only the EGFR has been shown to undergo ligand-induced internalization, although all members of the family display a significant rate of constitutive endocytosis (1–2%/min) . Members of the EGFR family interact extensively following ligand activation through heterodimerization, resulting in receptor transactivation . Overexpression of erbB-2, which is commonly observed in many human cancers, inhibits the lysosomal targeting and down-regulation of the EGFR . This appears to be due to competition between erbB-2 and the EGFR for components of the lysosomal targeting machinery . This results in enhanced receptor recycling, which could enhance signaling through the EGFR pathway.
Expression of the truncated, viral form of c-Cbl also enhances EGFR recycling, contributing to cell transformation . Genetic experiments indicate that sli-1, the Caenorhabditis elegans homologue of c-Cbl, negatively regulates the nematode EGFR, let-23 . c-Cbl is thought to stimulate EGFR degradation by acting as a ubiquitin-protein ligase, or E3 [26,27]. The PDGF receptor is also negatively regulated by c-Cbl , as is the CSF-1 receptor . Activation of the EGFR causes c-Cbl association within the endosome compartment where it is proposed to accelerate lysosomal targeting . Interestingly, c-Cbl associates with the CSF-1 receptor at the cell surface where it apparently functions in the acceleration of receptor internalization .
Leukowitz et al.  failed to observe a significant association between c-Cbl and erbB-2. It has also been shown that EGFR:erbB-2 heterodimerization inhibits c-Cbl binding . This is consistent with the observed inhibition of EGFR degradation following erbB-2 overexpression . It should be noted, however, that c-Cbl was found to stimulate erbB-2 degradation following its massive overexpression .
It has been suggested that c-Cbl is the primary regulator of EGFR trafficking between the early and the late endosomes , but this seems unlikely for a number of reasons. First, the activity of c-Cbl requires receptor kinase activity and phosphorylation at residue 1045 . However, numerous studies have shown that endosomal sorting and lysosomal targeting do not require receptor kinase activity [16,19,20]. Furthermore, EGFRs truncated to residue 1022 are internalized and degraded at a rate indistinguishable from full-length receptors .
Even though c-Cbl is unlikely to be a component of the endocytic sorting machinery, it does appear to be an important regulator of activated RTKs. Knock-out mice lacking c-Cbl show hyperproliferation and excess branching in the mammary epithelium, as would be expected of a negative regulator of internal EGFR pools . Overexpression of c-Cbl significantly stimulates the ligand-induced degradation of EGFRs, as well as of PDGF receptors, and truncated forms of Cbl can act as dominant-negative inhibitors of EGFR sorting . This strongly suggests that c-Cbl interacts with receptor sites that are crucial for normal receptor trafficking. Although RTK activity is not required for lysosomal targeting of the EGFR, it can enhance the process in full-length receptors , consistent with a role for c-Cbl as a modulator of EGFR degradation. A model that is compatible with most current data is that c-Cbl binds to kinase-active EGFR, mediates receptor ubiquitination, and then dissociates. Ubiquitin functions as a receptor ‘tag’ that increases receptor affinity for the lysosomal sorting machinery, resulting in enhanced receptor degradation. Alternatively, ubiquitinated receptors could be degraded by an alternate, proteosome-mediated pathway that does not involve lysosomes .
Molecules such as c-Cbl can clearly modify the lifetime of activated EGFR–ligand complexes, but how would this affect signaling? Many different studies have shown that internalized EGFRs are enzymatically active, hyperphosphorylated, and associated with Ras-GAP, Shc, Grb2, and mSOS [34,35]. The tyrosine kinase adaptor protein Shc seems to be strongly associated with active EGFRs at both the cell surface and during endocytic trafficking . This indicates that internalized EGFRs are capable of activating the same signaling pathways as surface-localized receptors. However, some studies have suggested that specific EGFR signaling pathways are triggered within endosomes .
Recently, several investigators exploited the differential pH-sensitive binding of transforming growth factor (TGF)-α and EGF to the EGFR to investigate endosomal signaling [36,37]. Because TGF-α binds poorly to EGFR within endosomes, the removal of surface-localized ligand with a mild acid strip results in the total loss of TGF-α-activated receptors. However, because EGF binds well to its receptor within endosomes, the removal of surface-localized EGF leaves a significant population of activated, endosome-localized receptors. Thus, it is possible to determine the specificity of endosomal signaling by comparing cells treated with TGF-α versus EGF following the removal of surface ligand. Such studies have shown that endosomally localized EGFR can activate Ras as efficiently as surface-localized receptors, but that activation of phospholipase C (PLC)-γ is restricted to the cell surface [37,38]. A similar approach has been used to demonstrate that internalized EGFRs are responsible for enhanced expression of the cell cycle regulatory protein p21/CIP1 . Interestingly, internalized EGFRs were unable to activate the PLC-γ pathway due to a lack of the appropriate lipid substrate in the endosomal compartment . Together, these results suggest that signaling from endosomes is qualitatively different from that generated at the cell surface. In addition, because EGFR heterodimer partners, such as erbB-2, remain on the cell surface following EGFR activation and internalization, it appears likely that heterodimer signaling remains restricted to the cell surface .
The biological significance of compartment-restricted signaling in the context of the EGFR system is unclear, but one possible function may be to discriminate between different EGFR ligands. There are currently six known ligands for the EGFR: EGF, TGF-α, amphiregulin, heparin-binding EGF-like growth factor (HB-EGF), betacellulin, and epiregulin . It has been shown that each of these ligands has distinct biological activities, but the mechanism responsible for their diverse actions is unknown . Biochemically, their ability to activate the EGFR appears to be identical. However, because TGF-α rapidly dissociates within the acidic environment of endosomes, its signaling pattern would be biased toward the cell surface . In contrast, ligands such as EGF that remain receptor-associated within endosomes would display a pattern that was biased toward the endosomes. For example, it is known that PLC-γ activation is necessary for EGFR-stimulated cell migration . The observation that TGF-α has a more pronounced effect on cell migration than EGF could be explained by the restriction of PLC-γ signaling to the cell surface. A role for compartment-specific signaling in ligand discrimination is especially plausible in the case of ligands that can function as either soluble or membrane-anchored growth factors. For example, membrane-anchored HB-EGF induces different biological responses than soluble HB-EGF .
The Insulin Receptor (IR)
The IR is an important metabolic regulator. Like the EGFR, the IR is a tyrosine kinase receptor, although it is structurally and functionally dissimilar . The dominant tyrosine kinase substrate for the IR is the insulin receptor substrate-1 (IRS-1) . Phosphorylated and activated IRS-1 recruits and activates a variety of signaling molecules, in a similar fashion to the phosphorylated carboxy terminal domain of the EGFR. IRS-1 is often found to be associated with cellular membranes, but it is neither inserted into the membrane nor stably associated with the IR itself . Although there are indications that IRs can signal from endosomes, most studies suggest that all or most signals are predominantly generated from the cell surface.
A functional role for signaling endosomes in IR signaling is supported primarily by temporal correlations. Insulin binding activates and induces rapid internalization of the IR . Studies in hepatoma cells, an in vivo rat model, and in adipocytes demonstrated that activated IRs accumulate in endosomal compartments in the presence of insulin . Although it has been suggested that as much as 75% of stimulated PI3K activity and 90% of phosphorylated IRS-1 is located in endosomal membranes , other studies suggest that the level of activated receptors in endosomes is low. After insulin-induced endocytosis, insulin and the IR promptly dissociate because their association is sensitive to the acidic pH found in the endosome . Dissociation of insulin from the IR allows receptor recycling and rapid insulin degradation. Evidence of insulin degradation by endosomal acid insulinase appears within 1 min . Mutations in insulin that stabilize its receptor association in acidic environments accelerate IR degradation and may be the cause of class 5, Type II diabetes . Consistent with a rapid dissociation and degradation of insulin, IRs located in endosomes are hypophosphorylated relative to surface receptors .
The dynamin mutant K44A has been used to determine the role of endocytosis in IR action . These studies have shown that the inhibition of IR endocytosis has no significant effect on either receptor or IRS-1 phosphorylation. In addition, activation of the serine, threonine kinase Akt is normal and insulin-stimulated DNA synthesis is unaffected. However, glucose transport, amino acid transport, glycogen synthesis, and lipogenesis are elevated, which is consistent with the hypothesis that blocking endocytosis leads to prolonged activation of the IR. Thus, the acute actions of insulin are largely independent of IR endocytosis and are initiated at the plasma membrane . Certain aspects of IR signaling were found to be attenuated, such as Shc phosphorylation and mitogen-activated protein kinase (MAPK) and PI3K activation, but there appeared to be no biological consequence of this attenuation. Interestingly, it has recently been shown that insulin and insulin-like growth factor (IGF-1) activation of the Shc and MAPK pathways in some cells is mediated through transactivation of the EGFR . Therefore, although it is unclear whether IRs directly generate signals from endosomes, signals could be generated indirectly through receptor transactivation.
Retrograde signaling mediated by the neurotrophin receptor TrkA may represent the best system supporting the idea that specific signals are generated by endosomes. Neurotrophins, such as nerve growth factor (NGF), stimulate the development and establishment of the peripheral nervous system . NGF stimulates neuronal cells through a family of receptors known by the archaic name of tropomyosin receptor kinase (Trk, A, B or C) . TrkA, like the EGFR, is an RTK. NGF binding induces Trk dimerization, autophosphorylation, and rapid endocytosis . Biochemically, the Trk receptor resembles the EGFR in many ways, and activates many of the same signaling partners and second messenger systems. Unlike EGFRs, however, TrkA signals must be propagated over great distances .
Neurotrophic factors bind their receptors at the sites of innervation near the axonal tip. These sites are sometimes at distances as great as 1 m from the neuronal cell bodies, and signals must reach the cell body to influence neuronal development and cell survival. It has been proposed that retrograde, vectoral transport of signaling endosomes from neuronal tips back to the cell body is the mechanism responsible for the long-distance transmission of neurotrophic signals. Thus, endocytic vesicles, formed at the axonal terminus and containing NGF–TrkA complexes, are thought to be transported back to the cell body along the axonal mictrotubule network where they generate the NGF signal .
Several lines of evidence suggest that the retrograde transport of endosomes is necessary for neurotrophic signal transmission. NGF is transported from axonal tips to cell bodies in the form of a complex with TrkA . The microtubule inhibitor colchicine inhibits the transmission of NGF signals, as well as the transport of NGF complexes . It has also been observed that internalized TrkA is tyrosine phosphorylated and is associated with the signaling partner PLC-γ. Interestingly, some evidence suggests that the transport vesicles containing NGF and activated TrkA are specialized. These vesicles, unlike coated or uncoated endocytic vesicles or synaptic vesicles, contain 10% of cell-bound NGF and almost one-third of activated TrkA . Characterization of these vesicles may aid efforts to better understand endosomal signal transduction.
It is difficult to study TrkA trafficking in neuronal cells because of the commonly present p75, which also binds NGF. To circumvent this problem, Zapf-Colby and Olefsky  used CHO cells stably transfected with TrkA. They found that TrkA was rapidly internalized upon binding NGF, but was poorly targeted to the lysosomes. Instead, most of the NGF was recycled intact back to the cell surface. NGF was bound to TrkA at endosomal pH values, suggesting that most of the internalized NGF was present as a ligand–receptor complex . The lack of efficient lysosomal targeting sequences in the TrkA receptor and the relative pH insensitivity of NGF binding strongly suggest that the TrkA–NGF system has evolved to promote endosomal signaling.
Strong support for the generation of specific signals from endosomes has recently been produced by Segal's group . Using a temperature-sensitive dynamin mutant, they demonstrated that cell surface Trk receptors promote NGF-induced survival while internalized receptors are more active in promoting differentiation. Internalization of activated Trk results in the termination of Akt signaling, but also causes a strong, transient level of Erk activation. Although a role for signal duration and magnitude in the generation of specific biological responses could not be excluded by their studies, the simplest explanation is that specific signal transduction pathways were triggered by internalized receptors.
In general, the studies described above provide strong support for the generation of specific signals from within endosomes. However, endosomal signaling is unlikely to be a universal aspect of RTK action. It could, however, serve specific roles, such as promoting neuronal differentiation. It is interesting that in the case of the IR system mechanisms, in the form of a pH-sensitive insulinase, appear to have evolved to prevent endosomal signaling . This leads to, perhaps, a more general rule: unless specific mechanisms exist to remove and degrade the ligand, signaling will continue after endocytosis. The IR system has probably evolved a rapid ligand degradation system to facilitate the timely termination of insulin action. The TrkA system has evolved to promote ligand stability and to slow receptor degradation. This is probably a necessary requirement for axonal transport. The EGFR system appears to have the greatest versatility. Some ligands, such as TGF-α, rapidly dissociate and are degraded within endosomes , while others, such as EGF, remain associated and are resistant to endosomal degradation. This ligand-specific processing and the consequent receptor trafficking may play an important role in generating different biological responses.
Clearly, the next steps that need to be taken are to understand the mechanistic basis of selective endosomal signaling and its biological consequence. What is the relative contribution of signal duration and specific substrate assembly? What determines where an RTK can productively interact with its signaling partners? Is it at the level of spatial segregation, as in the case of erbB-2, or is it restricted by the availability of substrates, such as PLC-γ? Answers to these questions will require methods to separate and analyze the composition of endosome signaling complexes. Also required is a better understanding of the link between substrate phosphorylation, induction kinetics, and biological responses.
The authors apologize to those whose work could not be cited due to space limitations. We are grateful to Doug Lauffenburger, Kevin Schooler, Alan Wells, and Becky Worthylake for many stimulating discussions and to Lee Opresko for critically reading the manuscript. This work was supported by NIH grant PO1-HD28528 (H.S.W.) and a predoctoral fellowship from the US Army Breast Cancer Research Program (P.M.B.).