Endocytosis and spatial restriction of cell signaling

Abstract

Endocytosis and recycling are essential components of the wiring enabling cells to perceive extracellular signals and transduce them in a temporally and spatially controlled fashion, directly influencing not only the duration and intensity of the signaling output, but also their correct location. Here, we will discuss key experimental evidence that support how different internalization routes, the generation of diverse endomembrane platforms, and cycles of internalization and recycling ensure polarized compartmentalization of signals, regulating a number of physiological and pathologically‐relevant processes in which the resolution of spatial information is vital for their execution.

1. Introduction

One essential feature of cells is their ability to respond to spatial information. This is generally achieved by adapting the architectural organization of their cytoskeletal and membrane components and more importantly signaling machineries so as to maintain an asymmetric and polarized distribution of molecules whose signaling output, thus, become spatially restricted. Accordingly, spatial restriction of signals has emerged as a critical device for the execution of a number of diverse and fundamental cellular processes, including directed cell migration, cell‐fate decision, epithelia‐cell polarization, growth cone movement, tissues morphogenesis during development, and cell invasion into the surrounding tissues of metastatic cells.

The textbook view of the key function of the endocytic process is to regulate nutrient internalization, signal transduction, and the composition of the plasma membrane. However, there are multiple mechanisms that can be envisioned through which the internalization and trafficking of membrane and membrane‐bound molecules may control the localized output of signaling cascades.

For instance, by efficient internalization and intracellular transport of plasma membrane proteins and their associated signaling molecules, the endocytic process is commonly regarded as a critical mean to attenuate extracellular ligand‐induced response. This is intuitively achieved by direct clearance of receptors, which function as the first‐line sensors of the cellular microenvironment, from the plasma membrane limiting the intensities of the ensuing signaling events. In addition to this, internalized activated receptors are often destined to trafficking routes that lead to their degradation, thus hampering the ability of cells to respond to the continuous presence of extracellular cues by directly reducing the number of responsive receptors.

A large body of emerging evidence, however, points to the fact that signaling is not restricted to the plasma membrane. As internalization proceeds, activated transmembrane molecules, with their tails exposed toward the cell cytoplasm, are confined into endomembrane organelles, which thus become bona fide signaling platform influencing not only the time and amplitude of the resulting signal, but also its specificity (Gould and Lippincott‐Schwartz, 2009). Consistent with this view, more and more signal transduction pathways are reported to require an active endocytic machinery, or strikingly to originate from various types of endosomes. A variation and an extension of this latter concept has further emerged with the recognition that plasma membrane receptors are internalized through different pathways, e.g. Clathrin‐mediated endocytosis (CME) or non‐Clathrin endocytosis (NCE), which have been shown to control directly the biological outcome of their signaling (Le Roy and Wrana, 2005). During the endocytic transport, molecules undergo a discrete set of route‐dependent posttranslational modifications, such as phosphorylation/dephosphorylation and ubiquitination, that directly influence the composition of the signaling cascade that is being activated (Sorkin and Goh, 2009). Thus, the biological output is controlled not only through compartmentalization of signaling into endosomal platforms, but also by the routes through which molecules reach the different compartments.

Endocytic internalization of membrane and membrane‐associated proteins is frequently accompanied by recycling of these factors back to the plasma membrane. This may function to replenish ligand‐free receptor for the next round of signaling and transport. Alternatively, the internalization/recycling cycle can also serve either as a mean to redirect and confine signaling molecules to specialized and distinct areas of the plasma membrane, such as the apical and basal membrane of polarized epithelial monolayer (Lecuit and Pilot, 2003), or as a positive feedback mechanism capable of maintaining the polarization state of critical signaling molecules, such the small GTPase Cdc42 during the formation of polarized buds in budding yeast (Marco et al., 2007).

Finally, endocytosis of molecules is accompanied by the internalization and recycling of plasma membrane generating a constant membrane flow. In analogy to the actin tread milling cycle, this flow of membrane had early been proposed to either generate forces that support the extension of migratory protrusions (Bretscher and Aguado‐Velasco, 1998b), or to promote the rearward movement of molecules bound to the surface of many cell types, as it may occur in cell protrusions of motile cells. Results consistent with membrane flow have been obtained from HeLa cells (Bretscher, 1983), fibroblasts (Schmoranzer et al., 2003), where membrane vesicles can be seen approaching and fusing with the leading edge. However, experiments with leukocytes and keratocytes failed to detect any significant rearward membrane flow (Kucik et al., 1990; Lee et al., 1990), suggesting that this is not a universal property of moving cells, although it may be important for some of them.

Recently, the requirement of a continuous flow of membranes propelled by key endocytic molecules, such as Clathrin, was shown to be essential for the extremely dynamic changes of cell shape that occur during directional, chemotactic migration of the amoeba Dyctiostelium discoideum, a professional mover (Traynor and Kay, 2007; Wessels et al., 2000). A similar role of CME in controlling cell shape is also at play in a variety of migratory cells, from mammals to zebrafish (Palamidessi et al., 2008), supporting the existence of a tight coupling between membrane dynamics, cell shape changes and polarized signaling.

Thus, endocytosis plays an important role in spatially‐restricting a diverse array of signaling outputs. In this review, we will cover prototypical examples of how different routes of internalization from the plasma membrane, the generation of endomembrane platforms, and cycles of internalization and recycling ensure spatial restriction of signals, regulating a variety of fundamental, physiological and in some cases pathologically relevant processes.

2. Signaling on the road: the internalization routes influences signaling output

A key event in the resolution of extracellular cues into a defined set of biochemical events controlling biological output is the binding of secreted ligands to plasma membrane receptors. These latter molecules are generally kept on idle until ligand binding, which triggers their activation at the plasma membrane and the ensuing assembly of macromolecular effector complexes required for signal propagation. In addition, activated receptors are concomitantly internalized entering trafficking routes leading either to their degradation or recycling back to the plasma membrane (Figure 1). The canonical view holds that the internalization process by clearing receptors from the plasma and promoting their degradation is a way to dampen continuous signaling. However, there is the possibility that signal transduction events may not be limited to the plasma membrane, taking place, instead, en route while the ligand/receptor complex traffics through distinct intracellular, endosomal compartments.

Figure 1.

Endosomal vesicles, trafficking routes and their regulators. Plasma membrane (PM) and membrane‐bound proteins are internalized through various routes. Here, are depicted Clathrin‐dependent (CME) and Clathrin‐independent (NCE) endocytosis. The latter one includes also raft‐mediated pathways. Both routes converge into Rab5‐positive early endosomes (EE), which represent the first endosomal sorting station. From EE cargos can be redirected through a fast recycling, Rab4‐dependent routes back to the plasma membrane, or enter, via Rab8, into a Rab11‐endocytic recycling compartment (ERC) before being retargeted to the PM (Zerial and McBride, 2001). Alternatively, cargo can traffic to a Rab7‐dependent lysosomal degradative route, by entering into multivesicular body/late endosome (MVB/LE) before being degraded into lysosome (Dautry‐Varsat, 1986). In addition to these “canonical routes”, cargos, such as MHC‐I or the interleukin receptor that do not enter through CME, can be sorted into Arf6‐positive recycling pathways (Arf6‐RE) and redelivered back to the PM (D'Souza‐Schorey and Chavrier, 2006; Donaldson, 2005). It is unclear what are the precise regulatory factors and intermediate endomembrane(s) that may connect EE and ERC to Arf6‐RE.

It has long been assumed that CME is the main pathway that is involved in the downregulation of cell‐surface receptors, such as EGFR (Conner and Schmid, 2003). However, several cell‐surface receptors, including the EGFR itself, are endocytosed also through non‐Clathrin pathways (Figure 1). One of the best‐characterized NCE route is mediated by lipid microdomains enriched in sphingomyelin and cholesterol, which correspond to lipid raft (Simons and Ikonen, 1997). These lipid microdomains may function as regulatory platforms for signaling. Lipid rafts that are enriched in caveolin‐1, an oligomeric protein that associates with cholesterol and sphingolipids, forming peculiar internalization structures, named caveolae, provide a case in point.

Among the various functions ascribed to caveolae, they are also proposed to sequester receptors on the plasma membrane by virtue of their extremely low mobility, thereby preventing their overactivation, such as in the case of two receptor tyrosine kinases (RTKs): Platelet‐derived growth factor receptor (PDGFR) and EGFR. Strikingly, cell stimulated with EGF no longer responded to PDGF, due to sequestration of the PDGFR into caveolae. Analogously, the same mechanism was observed in the case of EGFR in cells previously treated with PDGF (Matveev and Smart, 2002).

A role of lipid raft as a plasma membrane‐confined sink of signaling molecules has been recently extended also to other intracellular proteins, such the non‐receptor tyrosine kinase, Src. The binding of activated c‐Src to the raft‐anchored Csk adaptor, Cbp/PAG (Csk‐binding protein/Phosphoprotein Associated with Glycosphingolipid‐enriched microdomains), leads to the entrapment of the activated kinase into this microdomain, resulting in efficient suppression of c‐Src‐dependent signaling and Src‐dependent transformation (Oneyama et al., 2008; Veracini et al., 2008), pointing to a critical role of rafts in spatial‐restriction of signal in tumor development.

Numerous are the examples showing that the routes of internalization control the biological outcome, by determining the extent, duration and localization of downstream signaling (Sadowski et al., 2008). TGF‐β receptors can be internalized through both CME and NCE. When the adaptor, FYVE‐domain containing protein Sara (Smad anchor for receptor activation) binds to TGF‐β receptors at the plasma membrane, the receptor is routed to signaling‐competent endosomal vesicles (Di Guglielmo et al., 2003). Conversely, when Smad7 is recruited to the receptor, the E3 ubiquitin ligase Smurf is brought along, leading to ubiquitination of the receptor, its internalization through the raft routes, and its subsequent degradation (Di Guglielmo et al., 2003) (Figure 2). CME and NCE internalization pathways appear to be highly interconnected, and the decision as to whether take the signaling versus the degradation routes may be made not only at the plasma membrane, but also in intracellular sorting stations, such as the endosomes. Recently, for instance, the small GTPases Rap‐2 was shown to compete with Smad7 for binding the TGF‐β receptor. In the absence of ligand, Rap‐2 binding blocks receptor degradation in lysosomes and directs it to recycling through Rab11 recycling endosomes. In the presence of ligand, the receptor is also diverted from degradation, but instead of recycling, it signals (Choi et al., 2008). This event appears to occur after the initial steps of Clathrin‐ or non‐Clathrin‐dependent formation of internalized vesicles, suggesting that multiple check points or crossroads along the endocytic routes are in place controlling dynamically, in a ligand‐dependent manner, the way to take, ultimately influencing the final biological outcome.

Figure 2.

The internalization routes influence signaling output. (A) CME promotes sustained signaling, while NCE target cargos to degradation. EGFR and TGF‐β receptors enter cells either via CME or raft‐dependent, NCE pathways. EGFR and TGF‐β entering through CME are efficiently recycled. This trafficking route is coupled to sustained and spatially restricted signaling. Raft‐mediated internalization mainly targets these receptors to degradation leading to signal attenuation (Le Roy and Wrana, 2005). Ubiquitination of the receptors mediated by the E3 ligases Smad7 (Di Guglielmo et al., 2003) or CBL (Sorkin and Goh, 2009) for TGF‐β and EGFR, respectively preferentially direct these cargos toward degradative pathways, indicating that this posttranslational modification might be critical to determine receptor fate and signaling outcome, albeit the underlying molecular mechanisms responsible for this routing are ill‐defined. The cycle of internalization and recycling associated to CME is thought to be essential to spatially restrict signaling leading to directional migration and chemotaxis. (B) NCE promotes signal activation, while CME signal attenuation. The Wnt family of secreted ligands binds to the seven transmembrane‐spanning receptors of the Frizzled‐type and to the Low Density Receptor‐related protein (LRPs) (Kikuchi et al., 2007). Among Wnt ligands, Wnt3a can bind to its co‐receptor LRP6, promoting its raft‐mediated internalization and signaling. In the presence of Wnt3a, LRP6 is phosphorylated and internalized into a raft‐dependent, caveolin‐positive vesicular compartment, where it can stabilize β‐catenin and transduce the signal. Phosphorylation does not require the trafficking of the receptor to the caveolin‐1 positive compartment (where its kinase, CK1 g, is found) (Yamamoto et al., 2008). Conversely, signal transduction (i.e., β‐catenin stabilization) requires both phosphorylation and endocytosis, suggesting that endosomes may become critical signaling platform. LRP6 can also bind a Wnt3a antagonist, Dickkopf (Dkk), which diverts it from the caveolin to the Clathrin pathway, preventing the encounter of the receptor with its kinase, its phosphorylation and β‐catenin stabilization.

A somewhat similar scenario seems to operate also in the case of the signaling and internalization of the EGFR (Figure 2). In this latter case, the routes of internalization have been shown to be a function of ligand concentrations (Sigismund et al., 2005). When stimulated with low amounts of EGF, the receptors, at least in some cells, are internalized exclusively through CME (Sigismund et al., 2005). Conversely, at higher doses of EGF, a combination of CME and NCE mediates the internalization of the receptor (Sigismund et al., 2008, 2005). Under these conditions, the receptor becomes ubiquitinated and directed to a multi‐vesicular body‐dependent degradative pathway (Sigismund et al., 2005). It has to be pointed out that this scenario may be more complex since EGFR entry routes appear to be influenced not only by the doses of ligand, but also by the concentrations of the receptors on the cell surface (Sorkin and Goh, 2009). This notwithstanding, CME leads, at variance with respect to NCE, to recycling of the receptor, sustained signaling, while leaves degradation unaltered (Sigismund et al., 2008), as shown by the effects of blockade of CME on phosphorylation of EGFR signaling targets, such as AKT and ERK1/2. It is of note that the phosphorylation of other targets, such as Shc, does not require CME, suggesting that Clathrin endocytosis does not generally equal signaling, but rather activation of a particular subset of branches of the signaling network.

The mechanisms through which CME supports sustained EGFR signaling are yet to be fully established. One possibility is that CME biased receptors to enter a recycling pathway, allowing for multiple cycles of signaling at the plasma membrane or on endosomes, while diverting the receptors from degradation. This mechanism may be functionally relevant, for instance, when cells have to sense and respond to conditions of limiting ligands availability. An alternative, non‐exclusive possibility is that this internalization/recycling cycle may promote spatial‐restriction of signaling at the plasma membrane, which, as we will examine in more detail later on, may be critical for processes that require cell to become polarized, such as in directional migration.

One additional levels of complexity is that not all EGFR entering via CME is recycled, but a sizable amount is destined to degradation (Sigismund et al., 2008) (Figure 2). This raises the issues as to whether the decision to take different fates (recycling versus degradation) is stochastic or underscores a precise molecular mechanism. This latter possibility is indirectly supported by evidence showing that coated vesicles have a variable composition and may be formed by a different repertoire of molecules that marks cargo and machinery‐specific CME routes (Sorkin, 2004). In the case of EGFR, one may assume that the receptor when stimulated at low doses of ligand, ends up in Clathrin pits or coats of variable composition, despite being internalized through the same CME pathway. These Clathrin structures might be composed of different accessory proteins that influence the fate of cargos to recycling and signaling or to degradation. Indeed, ablation of the major Clathrin adaptor AP‐2, blocks by about 50% CME‐dependent EGFR internalization (at low doses of ligand) (Sigismund et al., 2008), suggesting that non‐AP‐2‐containing pits do exist. It is also remarkable, that the perturbation induced by AP‐2 interference on recycling and signaling are similar to the one obtained by complete blockade of CME, indicating that AP‐2 marks specific recycling and signaling routes, as oppose to degradative pathways (Sigismund et al., 2008). Finally, it is worth to point out that EGFR, independently of its entry routes, is funneled to early endosome whose composition appears to be identical (EEA‐1, Rab‐5 positive), yet cargo are differentially sorted, like they remember the way they entered the cells. One simple molecular explanation accounting for this acquired “memory” is that different posttranslational modifications of the activated receptors (differential number of phosphorylated tyrosines or of appended ubiquitin molecules) might occur as function of ligand concentrations. While this is subject of heated debates (Sorkin and Goh, 2009), it is intriguing that ubiquitinated EGFR could be revealed only after addition of a discrete concentration of EGF, suggesting the existence of a threshold, which must be reached for the EGFR to be modified by ubiquitin and enter a non‐Clathrin pathway (S. Sigismund, personal communication). It remains unclear, instead, why and how differentially modified cargo might be first sorted at the plasma membrane, in a ligand‐dependent manner, to be remixed in early endosomes, where additional sorting events (recycling versus degradation) take place.

The dependency from ligand concentrations of the fate of internalized receptors may have important and physiological relevant implications. A number of plasma membrane receptors, including RTKs and G protein couple receptor (GPCR), function as motogenic sensors, which especially during developmental morphogenesis, respond to gradients of chemotactic factors that guide the migration of cells to their final destination. These motile cells, in addition to migrate directionally, must also be capable to arrest at their target sites, where the concentration of the chemotactic factors is the highest. Thus, a ligand‐dependent internalization/sorting mechanism that drives a motogenic receptor toward degradative pathways may be critical to switch off the migratory signal when appropriate. A scenario of this kind has been for instance demonstrated to occur during the migration of primordial germ cells (PGC) toward the gonads in zebrafish development (Raz, 2004). PGCs express the chemokine receptor CXCR4b and directionally migrate toward sites in the embryo at which the ligand SDF‐1a is expressed (Raz, 2004). Binding of SDF‐1a elicits also the internalization of the receptor promoting its spatial redistribution and restricting its signaling, two events that are required for proper chemotaxis. Indeed, an internalization‐defective receptor led to aberrantly elevated signals, increased time spent “running”, preventing cells to reach their final target, while promoting ectopic cell migration (Minina et al., 2007).

Alternatively, the concentration of ligand may determine not only the duration or intensities of the response, but also its specificity. In keeping with this latter notion, the dependence on ligand concentration has also been observed for the PDGF and its cognate receptor. However, in this case, cells stimulated with low doses of PDGF responded by rearranging their cytoskeleton and acquiring migratory properties in a CME‐dependent manner. Mitogenesis and cell proliferation were, instead, preferentially induced by higher concentration of PDGF, a response that required a functional raft/caveolin pathway (De Donatis et al., 2008).

Both the EGF and TGF‐β receptor systems utilize the CME for signaling/recycling, and the NCE for degradation. Recently, it was shown that other cargos exploit the different internalization pathway in the opposite manner. This is the case of the Wnt3a‐activated low‐density receptor‐related protein 6 (LRP6) (Figure 2). The Wnt family of secreted signaling molecules is essential in embryonic development, cell polarity generation, and cell fate specification (Mikels and Nusse, 2006). The canonical Wnt pathway involves activation of β‐catenin‐dependent transcription and is evolutionarily conserved from Caenorhabditis elegans to humans. Wnt molecules bind to two coreceptors, the Frizzled‐type seven‐transmembrane‐domain receptor and the LRP5/6 (Kikuchi et al., 2007). Yamamoto et al. (2008) recently showed that the Wnt3a signal activation by its co‐receptor LRP6 happens in the raft route, while signal attenuation is associated to the Clathrin route (Figure 2). In the presence of Wnt3a, LRP6 is phosphorylated and internalized into a raft‐dependent, caveolin‐positive vesicular compartment, where it can stabilize β‐catenin and transduce the signal. Phosphorylation does not require the trafficking of the receptor to the caveolin‐1 positive compartment (where its kinase, CK1 g, is found). Conversely, signal transduction (i.e., β‐catenin stabilization) requires both phosphorylation and endocytosis, suggesting that endosomes may become critical signaling platforms. LRP6 can also bind a Wnt3a antagonist, Dickkopf (Dkk), which target the receptor to CME. However, signal transduction does not happen along this route, which prevents the encounter of the receptors with its kinase CK1 g (Figure 2).

Thus, CME and NCE are differentially coupled to signal activation or attenuation depending not only of the types of cargo, but also on their mode of activation and internalization promoted by different ligands or different concentrations of the same ligand. While this prevents to propose a general theory connecting signaling and trafficking, it highlights how animal cells can increase their ability to plastically adapt their response to extracellular cues by exploiting the diversity of their trafficking networks.

3. Endosome as signaling platforms

After the initial internalization events that include the invagination of the plasma membrane, the scission of the resulting tubular structures, and the formation of a coated, in the case of CME, or choleresterol‐rich vesicle, in NCE, cargos are invariably destined to early endosomes (Figure 1). These vesicular compartments are characterized by a reduced pH (around 6), and a well‐defined and peculiar lipid and protein composition, which typically include the small GTPase Rab5 and the early endosome antigen‐1 (EEA‐1) (Zerial and McBride, 2001). Early endosomes represent the first sorting station in the internalization process. From these compartments cargos are either directed to multi‐vesicular body (MVB)/late endosome and lysosome for degradation or recycled back to the plasma membrane (Figure 1). Increasing evidence has been accumulating supporting the hypothesis that endosome are signaling platforms, where signals are originated and kept compartmentalized so as to generate quantitative and qualitative differences, thus serving as an effective way to restrict signals in a spatially and temporally controlled fashion.

The “signaling endosome hypothesis” was originally proposed in neurons where it is most obvious the requirement for a long‐range transmission of signals [for review see ref. Howe and Mobley (2005)]. Neurotrophin receptors at nerve terminals must send their signals controlling events in the neuronal body and the nucleus that can be positioned at long distance (up to 1m) from the terminal. The speed and distance at which signals are sent cannot be accounted by simple diffusion. Indeed, we now know that internalized NGF is confined into early endosomes, where it remains bound to its cognate TrkA receptor, which is fully competent to signal to ERK1/2 and PI3 K (Phosphatidylinositol‐3‐kinase)/AKT. These endomembranes can be transported by retrograde microtubule‐and dynein‐based transport to their targets where signal is delivered promoting survival (Howe and Mobley, 2005).

Neurobiological studies have also provided evidence that different signaling events elicited in different compartments might trigger different biological outcomes. Thus, for example, whereas increased brain‐derived neurotrophic factor (BDNF) in the target of retinal ganglion cells promotes arborisation of both their axons and dendrites, increased BDNF within the retina itself inhibits dendritic arborisation (Lom et al., 2002). In the cerebellar granule cell, it has been shown that stimulation of growth cones with the motogenic guidance cue, Slit‐2, triggers growth cone collapse locally, but migration reversal distally, after retrograde propagation of a Ca2+ wave to cell bodies (Guan et al., 2007). Notably, the Ca2+ wave is by itself capable of inducing only migration reversal, and not growth cone collapse, indicating that Slit‐2 produces different effects in different compartments.

A large body of evidence that endosomal signaling is not restricted to neurons has accumulated in recent years thanks to the exponential progresses in live imaging coupled with the development of signaling biosensors. Several imaging methods have revealed activated RTK on endosomes. EGF, for instance, remains associated to its receptor in early endosomes as evidenced by the use of anti‐phosphospecific antibodies (Burke et al., 2001), or the presence in these vesicles of complexes, which are commonly bound to activated EGFR at the plasma membrane, mediating the RAS/ERK signaling cascade (Baass et al., 1995). More recently, Sorkin et al. (2000) were able to show directly the activation of EGFR along the endocytic route through the use of an FRET pair biosensor composed of a CFP‐tagged EGFR and an YFP‐tagged, Grb2. Similar conclusions were also reached using fluorescence lifetime imaging of GFP‐EGFR and microinjection of anti‐phosphotysine antibodies conjugated with an FRET acceptor (Wouters and Bastiaens, 1999).

In addition to direct visualization of activated RTKs also their targets, such H‐Ras, have been found to signal from early endosomes by live cell imaging of specific biosensors (Jiang and Sorkin, 2002). Notably, not all Ras family members display a similar endocytic requirement for signaling. For instance, H‐Ras, but not K‐Ras, activation was reported to be entirely dependent on functional endocytosis and to correlate with the accumulation in Rab5‐positive macroendosome of GFP‐H‐Ras (Parton and Hancock, 2004), suggesting that endocytic routes and endosomal localization may spatially restrict the signaling outputs of a subset of specific Ras variants. In keeping with and extending this notion, H‐Ras was recently shown to accumulate also in NCE‐derived, Arf6‐dependent macropinosomes, whose lipid and proteins composition significantly differs from that of early endosomes, suggesting that these stations may provide discrete intracellular microenvironments for specificity of H‐Ras signal (Porat‐Shliom et al., 2008).

These latter results also point to another important and likely more general mechanism through which endosomal vesicles may provide spatial regulation of signaling output: the heterogeneity of the endosomal compartments. In addition to macropinosomes, separate subpopulations of this early endocytic compartments also exist. Endosomes bearing the Rab5 effectors APPL1 and APPL2, for instance, could be distinguished from EEA‐1‐positive early endosomes and shown to be specifically required for mitogenic signaling (Miaczynska et al., 2004) (see also Pyrzynska et al., 2009). Remarkably, studies in Zebrafish demonstrated that APPL1 localized to endosomes not only regulates AKT activity, but also defines its substrate specificity, controlling GSK‐3β, but not TSC2, a necessary step to support cell survival during development (Schenck et al., 2008). Collectively, these studies indicate that several endosome populations can act as distinct platforms for assembly of different sets of signaling effectors. Whether EEA1 and APPL endosome are entirely distinct sorting stations or functionally linked trafficking intermediates has remained unclear. Recently, however, APPL‐endosomes have been shown to represent an early stage of the Rab5‐positive compartment, whose maturation is controlled by the localized production of different phosphoinostide (Zoncu et al., 2009). Imaging analysis of endocytic organelles have shown that newly formed CME or NCE‐derived vesicles mature into APPL‐1 or Rab5‐postive endosomes. As APPL1 endosomes move centripetally and phosphoinositol‐3‐phosphate (PI3P) is generated, APPL1 is shed and replaced by PI3P binding proteins, such as EEA1, revealing an unexpected level of plasticity of early endosome (Zoncu et al., 2009). Additionally, since APPL1 endosomes appear to be the one where “proliferative signal” originates, this finding implies that a PI3P switch, leading to endosome maturation may be critical also to control signaling strength.

In few cases, these endocytic structures have been shown to function as an obligatory intermediate signaling station between the plasma membrane and the nucleus. For instance, in response to EGFR stimulation, APPL1 translocates from these endomembranes to the nucleus, where it interacts with the nucleosome remodelling and histone deacetylase multiprotein complex NuRD/MeCP1, an established regulator of chromatin structure and gene expression (Miaczynska et al., 2004). Notably, the interaction of APPL1 with Rab5 appears to be part of a control mechanism to couple the release of APPL1 from the endosomes to growth factor signaling and trafficking (Miaczynska et al., 2004). Indeed, continuous rounds of binding of APPL1 to Rab5 and dissociation from the membrane may be necessary to ensure the reversibility of interactions with its downstream factors, but also to regulate their activity. While the molecular underlying nature of these control mechanisms remains elusive, it is plausible that posttranslational modifications or conformational changes taking place on the Rab5‐positive membranes may regulate the ability of APPL1 to functionally interact with other partners, making cycling through endosomes an obligatory step in APPL signaling.

A somewhat similar mechanism appears to be at play in the propagation of signaling from the tyrosine kinase receptor c‐Met to the activation of STAT3 (signal transducer and activator of transcription). Hepatocyte growth factor (HGF), also known as scatter factor, binds to its receptor, the c‐Met tyrosine kinase, to induce cell proliferation, migration, morphogenesis, and survival as the result of diverse signaling pathways (Birchmeier et al., 2003). How the different pathways downstream of HGF/c‐Met are specifically regulated and coordinated is still unclear. One downstream signaling molecule of c‐Met is the multifunctional transcription factor STAT3. Binding of HGF to c‐Met results in recruitment of STAT3 to c‐Met, phosphorylation of STAT3, and STAT3 nuclear accumulation (Birchmeier et al., 2003), a process shown to be dependent, at least in part, from endosomal localization of this transducer (Bild et al., 2002). In keeping with this latter notion, recent observations demonstrated an unexpected relationship between the strength of the signaling response (e.g., activation by phosphorylation) and the trafficking of the receptor and downstream signaling components (Kermorgant and Parker, 2008). By comparing the activation of ERK1/2 and STAT3 with the intracellular trafficking of c‐Met in response to HGF stimulation, it became apparent that HGF elicits a potent activation of ERK1/2 that requires internalization of c‐Met into endosomes, but does not require its trafficking to a perinuclear compartment. Conversely, STAT3 activation is relatively “weak” and requires the localization of active c‐Met to the perinuclear region via a microtubule and PKCα‐dependent process. However, STAT3 activation via the cytokine Oncostatin M produces a “stronger” signal that occurs independently of microtubules and PKCα and may likely be independent from functional endocytosis. This latter result demonstrates that the trafficking of STAT3 to endosomes is not an absolute requirement for activity, but rather depends on the “signal strength” elicited by the growth factor–receptor complex.

There are a number of questions raised by this model. For instance, it is unclear what are the molecular determinants dictating a “weak” versus a “robust” response to c‐Met activation. In analogy to the EGFR system, it would be interesting testing whether the ligand dose and timing of stimulation affect STAT3 response and how this correlates with the internalization routes of the receptor, and the peculiar microenvironment of the endosomes. Possibly, weakly activated STAT3 may survive longer in this microenvironment as it may be protected from cytosolically abundant phosphatases, and it may simultaneously be efficiently transported in a directional‐fashion toward the nucleus, where STAT3 biological action is primarily exerted. Another issue that remains to be addressed is the nature of the endosomal compartments where STAT3 signal is kept sustained. The perinuclear localization of the STAT3‐positive endomembrane is suggestive of late endosomal/recycling structures. However, in its trafficking toward the nucleus, STAT3 passes through EEA‐1 early endosomes and it is unclear whether these are key stations for the maintenance of its signal strength. As more and more receptors are shown to utilize this specialized APPL compartment [reviewed in Sadowski et al. (2008)], it is conceivable that also c‐Met signals through these intermediate structures. Thus, it will be worth testing whether STAT3 an APPL resides in the same endosome, raising the possibility that there might be a compartment where endocytic molecules exerting a nuclear function are specifically sorted.

The trafficking from early endosome to the lysosomal degradative compartment reaches another station: the late endosome or MVB (Figure 1). MVB are a morphologically distinct organelle marked by the presence of Rab7 and the lysosomal‐associated membrane proteins‐1 (LAMP‐1), and characterized by the formation of intraluminal vesicles where receptors are engulfed, isolated from their cytoplasmic effectors, and thus rendered signal incompetent (Hurley, 2008). MVB are, therefore, regarded as place of signal attenuation. In same cases, however, they have been shown to contain activated receptors and to promote ERK activation through compartment‐specific adaptors. This is the case, for instance, of the late endosomal adaptor p14, which in complex with MP1 serves as a scaffold recruiting MEK1 at the membrane thus leading to ERK activation (Teis et al., 2002). Notably, the same complex is incapable to promote ERK activation when p14 is mis‐targeted to the plasma membrane (Teis et al., 2002), strengthening the notion that endocytic organelle provide specific microenvironmental requirement for signaling. Another example of “signaling MVB” is found in neurotrophin‐mediated activation of late endosomal TrkA. At this site, TrkA is bound to the guanine nucleotide exchange factor for Rap1, PDZ‐GEF‐1, thus mediating the activation of this GTPase and ensuing ERK signaling required for neurite outgrowth (Hisata et al., 2007). In keeping with this notion, Rab7‐positive endosomes have also been implicated as the long‐range signaling carriers during the retrograde transport of neurotrophin (Deinhardt et al., 2006).

Collectively, these examples provide evidence that different endocytic compartments define spatially confined intracellular platforms controlling both signal intensity and specificity, thus providing an additional mechanisms to fine tune the regulation of a diverse array of biological outputs in response to extracellular stimuli.

4. Endocytosis and recycling spatially restrict signals controlling proliferation and cell fate determination, actin‐based cell migration and cell‐adhesion

The fate of internalized plasma membrane and plasma membrane‐bound proteins that are not directed to degradative endolysosomal pathway is to be recycled to their site of origin. The endosomal recycling pathway (Figure 1) is comprised of several types of endosomes, including early/sorting endosomes (EE), the endocytic recycling compartment (ERC), which is generally located at the cell center, near the Golgi apparatus (GA), and the Arf6 (ADP‐ribosylation factor‐6)‐dependent recycling compartments (ARC) that contains integral membrane proteins that are endocytosed into cells independently of AP‐2 and Clathrin (Donaldson and Honda, 2005). In addition to Arf6, many regulatory components of the endosomal pathway have been identified, including a number of Rabs, small GTPases that regulate distinct steps in the intracellular membrane pathways (Zerial and McBride, 2001) (Figure 1). Thus, membrane‐bound molecules are continuously subjected to an apparently cumbersome, energy‐expansive, futile cycles of removal from the plasma membrane to be replenished at the same sites after execution of their signaling function.

However, there are numerous ways and an increasing number of examples indicating that endocytic/recycling of membrane and membrane‐bound proteins is a process intimately intertwined with signaling, controlling not only its intensity, but also its specificity by spatially confining where signaling output occurs. Indeed, spatial restriction of signaling is emerging as an evolutionary conserved, critical device essential for the execution of a diverse set of fundamental cellular programs. Virtually all these processes are characterized by some sort of “polarized and/or asymmetric” distribution of molecules and functional activities, such as localized desensitization/resensitisazion in GPCR signaling, cell fate determination in asymmetric cell division, directed cell motility or the establishment and maintenance of epithelial tissue architecture. In the following section, we will analyze prototypical cases that demonstrate how endocytic/recycling, by ensuring localized intracellular response to various stimuli, promotes the spatial restriction of polarized signals.

4.1. Recycling resensitizes what endocytosis desensitized specifying signaling output

One obvious functional role of the endocytic/recyling routes is to control the levels and activity of plasma membrane receptors, such as in the case of the G protein–coupled receptors (GPCRs) (Hanyaloglu and von Zastrow, 2008). Studies of several GPCRs established the paradigm of rapid desensitization mediated by the coupling of receptor phosphorylation and internalization. According to this paradigm, agonist‐bound receptors initiate signaling by activating heterotrimeric G proteins at the plasma membrane and then rapidly undergo phosphorylation by GPCR kinases (GRKs) that phosphorylate agonist‐activated receptors (Hanyaloglu and von Zastrow, 2008). Phosphorylation of the receptor, and subsequent binding of the adaptor proteins β‐arrestin‐1 and β‐arrestin‐2 impedes the interaction of the receptor with G proteins, effectively terminating G protein‐mediated signal (Hanyaloglu and von Zastrow, 2008). Arrestins can also bind to Clathrin, thereby promoting CME of arrestin‐bound receptors, contributing to their desensitization. Sorting of internalized receptors into a rapid recycling pathway, by contrast, promotes the return of intact receptors to the plasma membrane and effectively re‐sensitize cells to respond again to extracellular ligands (Hanyaloglu and von Zastrow, 2008).

This cycle of desensitization/resensitization has important implications not only as a tightly regulated on/off switch of the signaling cascade that is thus rendered exquisitely sensitive to changes in extracellular ligand concentrations, but also in specifying the nature of the resulting signals. This latter notion is exemplified by a series of studies on β‐Adrenergic receptors (βARs). β‐Adrenergic receptors play critical roles in mediating physiologic responses to the hormone epinephrine and the neurotransmitter norepinephrine in animal hearts (Hanyaloglu and von Zastrow, 2008). Both β1AR and β2AR are expressed in cardiac myocytes and respond to the same stimuli, but they possess distinct functions in vivo. For instance, β1AR and β2AR display different trafficking and signaling properties in neonatal myocytes. Activated β1ARs remain on the cell surface and couple to Gs (Xiang et al., 2002). In contrast, activated β2ARs undergo robust endocytosis, and the receptors couple sequentially to Gs and Gi in cardiac myocytes (Xiang and Kobilka, 2003). The binding of the receptors to PDZ‐adaptors is critical in regulating these different endocytic and functional fates. Consistently, a mutation disrupting the interaction between the β1AR PDZ motif and its binding partners enables agonist‐induced receptor internalization (Xiang et al., 2002). The same mutation also enables the receptor to couple sequentially to Gs and Gi after activation (Xiang et al., 2002), virtually transforming a β1AR into a β2AR. Remarkably, this point mutation or addition of membrane‐permeable peptide that block the binding to PDZ domain‐containing proteins disrupted the receptor recycling and also inhibited receptor coupling to Gi, thus affecting the ability of the β2‐adrenergic receptor to “switch” G protein coupling specificity between distinct Gs‐ to Gi‐mediated pathways, which in turn are implicated, respectively, in acute regulation of cardiac contraction and long‐term control of myocardial cell survival (Hanyaloglu and von Zastrow, 2008). Thus, endocytic sorting of GPCRs, in addition to dynamically controlling surface receptor number, is intricately involved in cell signaling regulation and functional specification.

4.2. Sorting cell fate in asymmetric cell division

Compelling evidence for a critical role of endocytic/recyling in the determination of cell fate and control of asymmetric cell division comes from studies of the developing sensory organ in Drosophila, a process where endocytosis exerts a critical control at multiple levels (Coumailleau and Gonzalez‐Gaitan, 2008). Similar mechanisms of action have also been shown to operate in other model organisms and mammalian cells (Gonczy, 2008) (see for details Fürthauer and González‐Gaitán, 2009). The notion emerging from all these studies is that one general mechanism to achieve asymmetric signaling leading to cell fate determination is via polarized or biased endocytic/recycling events. It is remarkable that different steps of internalization are specifically employed to achieve spatial restriction of signaling, including differential downregulation by internalization of receptors function, selective endocytic activation of the cognate ligands and differentially regulated recycling. This suggests that specific regulation of distinct trafficking steps may provide a general, but highly versatile mechanism to ensure polarized signaling in cell‐ and tissues‐specific context, so as to control asymmetric division and self‐renewal programs.

4.3. Orchestration of cell motility

One process that depends on the intertwined connection between signaling and localization is cell motility, where the precise perception of extracellular cues in three‐dimensional setting is remarkably complex, particularly when cells move toward chemo‐attractants in a polarized fashion. Under these conditions, cells must reorient directionally by polarizing key plasma membrane proteins according to the direction of travel. Additionally, coordination between membrane traffic, cell substrate adhesion, and actin remodelling is required to generate propulsive forces responsible for the protrusive activity at the leading edge of motile cells.

4.4. Endocytosis and recycling promote directed cell motility

One mechanism that has been proposed to orchestrate protrusions in directional moving cells is the maintenance of the polarized state of critical motogenic sensors and their signaling effectors through a flow of endocytic internalization of membranes accompanied by redelivery of vesicles and cargos to the cell front (Bretscher and Aguado‐Velasco, 1998b). Originally, this idea was prompted by the observations that vesicles‐containing recycling cargos, such as the receptors for low density lipoprotein (LDL) and transferrin, are concentrated toward the periphery of the cells (Bretscher, 1983). To explain these non‐uniform distributions, the recycling of receptors was invoked as a way to return them to the cell surface at the cell's leading edge. Since the distribution of Clathrin‐coated pits on cells is uniform, Bretscher and Thomson (Bretscher, 1983) proposed that there is a bulk membrane flow toward the cell center that drive cell locomotion. This possibility was further supported by the finding that EGF stimulation to induce prominent ruffling of advancing plasma membrane required exocytic delivery of recycling membrane into just those specific sites where ruffles form (Bretscher and Aguado‐Velasco, 1998b). While bulk membrane flow as a mechanisms to propel cell motility has been questioned by experiments using single particle tracking to examine the movements of diffusing particles on rapidly locomoting fish keratocytes where the membrane current is likely to be greatest (Kucik et al., 1990), the possibility remains that polarized endocytic/recycling may act, at least in some cell types, as a device to confine cell signaling molecules, thus spatially regulating where the dynamic actin polymerization generating protrusive forces are exerted. Consistently, studies in genetically tractable model organisms provided the unequivocal evidence that endocytic/recycling are essential for cell directional migration in response to chemotactic gradient. In Drosophila melanogaster, the migration of border cells is regulated by gradients of motogenic growth factors, such as EGF and PV (PDGF/VEGF), toward which cells directionally move (McDonald et al., 2006). Genetic removal of CBL, a prototypical E3 endocytic ligase, or Sprint, a Drosophila homologue of the Rab5 GEFs Rin1, or the expression of a dominant negative Shibire/Dynamin mutant impaired in GTP hydrolysis and thus unable to promote vesicle scission, completely disrupted border cell migration (Jekely et al., 2005). Thus, endocytic pathways, particular those impinging on Rab5, are required to ensure spatial resolution of chemotactic signaling emanating from different RTKs, regulating actin‐based, polarized protrusive activity and motility. It has to be pointed out that genetic evidence in support of the requirement of polarized recycling for Drosophila border cells to migrate directionally has not yet been provided. However, there are several studies in mammalian cells that support a specific role of endosomal recycling in cell migration. For example, it has been shown that in KB cells, surface ruffles may arise from exocytosis of internal membrane from endosomal cycles (Bretscher and Aguado‐Velasco, 1998a), a process that is coupled to the polarized distribution toward the cell front of recycling membrane receptors (Bretscher and Aguado‐Velasco, 1998a). There is also some evidence that inhibition of the slow recycling pathway by expressing dominant negative Rab11 or the truncated Myosin Vb or Rab11‐FIP, an effector of Rab11, impaired HeLa cell migration (Mammoto et al., 1999) and chemotaxis of basophilic leukemia (RBL)‐2H3 cells (Fan et al., 2004). These latter results have recently been extended also in mammalian epithelial PtK1 cells, where, however, the interference with Rab11 recycling pathway increased random motility possibly as a consequence of delocalized formation of protrusive lamellipodia, but concomitantly impaired directional and persistent migration (Prigozhina and Waterman‐Storer, 2006). Thus, polarized endosomal recycling may not be required for cell locomotion per se, but is critical for the maintenance of cell migration polarity, which when disrupted leads to disorganized motility.

4.5. Endocytosis and recycling ensure spatial restriction of signaling controlling migratory programs in three dimensions

Similar mechanisms as the one describe above have also been proposed to account for the generation of different kinds of migratory protrusions induced after stimulation with RTKs, such as peripheral lamellipodia (PL) and dorsal surface circular dorsal ruffles (CDR) [reviewed in Buccione et al. (2004)] (Figure 3). Lamellipodia are the first obligatory step in two‐dimensional cell motility. The function of circular dorsal ruffles is less established, but has been shown to be sites of internalization events, such as fluid phase endocytosis (Buccione et al., 2004), and required for migration in three‐dimensional matrix (Suetsugu et al., 2003). These latter combined features suggest that circular ruffles may be sites where integration between actin dynamics and endocytosis take place. In keeping with this notion, initial dissection of the molecular machinery responsible for the formation of these protrusions revealed that endocytic proteins, such as Dynamin, and actin dynamics regulators, such as cortactin, function coordinately in the generation of CDR (also called wave) (Krueger et al., 2003; Orth and McNiven, 2003). Moreover, a tripartite signaling cascade, originating from Rab5, PI3 K, and Rac must simultaneously operate to allow the formation of circular ruffles (Lanzetti et al., 2004). Conversely, a pathway, linearly connecting, Ras‐PI3 K‐Rac to actin remodelling is sufficient to generate lamellipodia protrusions in response to RTK activation (Lanzetti et al., 2004). How these three‐pronged signaling pathways cross talk and are co‐regulated in space and time had remained unexplored. Recently, however, we showed that Rab5 endocytic/recycling trafficking of Rac is specifically required for the generation of CDR (Palamidessi et al., 2008). Consistently, inhibition of endocytosis in HeLa cells by depletion of Clathrin‐coated pits or expression of a GTPase‐deficient form of Dynamin inhibited Rab5‐dependent Rac activation and CDR formation. Additionally, Rac trafficking from the plasma membrane to endocytic vesicles containing Rab5 and Tiam1 were required for spatially‐confined Rac activation, although it remains unclear whether the active Rac on Rab5 endosomes comes exclusively from cell surface‐internalized Rac or whether there is also recruitment of cytoplasmic Rac from Rac–GDI (GDP dissociation inhibitor) complexes. This notwithstanding, to enable localized execution of Rac function, the internal pool of activated Rac GTP must be transported back to the plasma membrane to specific sites of “high actin activity”. Blockade of recycling by temperature switch to 16°C completely abrogated the formation of CDR in response to Rab5 activation and HGF stimulation. Furthermore, by exploiting Rac‐fused to photoactivable GFP (paGFP) coupled with two‐photon activation excitation that allows targeting of single internalized vesicles, confining the activation of paGFP molecules in three dimensions, while completely excluding the plasma membrane compartment, a direct visualization of Rac polarized transport toward ruffles in the plasma membrane could be evidenced in real time. The targeting of vesicles back to the plasma membrane is accomplished by the small GTPase Arf6, but not by Rab4 or Rab11 (Figure 3), indicating that the ARF6‐recycling compartment is the one utilized by Rac for its redelivery to the plasma membrane. Notably, Arf6 has previously been implicated in Rac trafficking (Radhakrishna et al., 1999). Moreover, Arf6‐mediated trafficking of vesicles is known to be responsible for recycling of certain receptors, such as the β1 integrins, to the plasma membrane (D'Souza‐Schorey and Chavrier, 2006), suggesting the intriguing possibility that these vesicles may be the site where different kinds of cargo are sorted and coordinately delivered to ensure spatial restriction of signaling required for polarized motility.

Figure 3.

Endocytic trafficking of Rac and integrin is required for spatial restriction of signaling in the control of migratory programs. (A) Hepatocyte growth factor (HGF) receptor activation induces the formation of circular dorsal ruffles (CDR, arrow) and peripheral lamellipodia (PL, arrowhead). This is accompanied by Clathrin‐mediated internalization of the cognate receptor tyrosine kinase, c‐Met. Clathrin‐ and Rab5‐dependent endocytosis is required for Tiam‐1‐mediated, a guanine‐nucleotide exchange factor, Rac activation in endosomal vesicles. Recycling of Rac back to plasma membrane areas is essential to promote the reorganization of actin into CDR. The small GTPase Arf6 targets activated Rac to specialized plasma membrane areas (such as ruffles) with high actin activity. The integrin α5β1 can associate with the small GTPase Rab25, a member of the Rab11 family, involved in vesicle recycling (Caswell et al., 2007). Both Rab25 and α5β1 integrin co‐localize in intracellular vesicular compartments within the distal tips of pseudopods (the lamellipodia equivalent in a 3D setting) (Caswell et al., 2007). Integrin α5β1 can traffic bidirectionally between intracellular Rab25 vesicles and the plasma membrane within the confines of the pseudopodial tips, this promotes the compartmentalization of a spatially restricted subpopulation of cycling α5β1 within the tip regions of extending pseudopods (Caswell et al., 2007). A HeLa cell stimulated with HGF and stained with phalloidin to detect F‐actin (white) is shown. (B) Mode of motility of individual cells in 3D. Two modes of single‐cell 3D migration in extracellular matrices have been described (Wolf and Friedl, 2006). The elongated‐lamellipodia movement (mesenchymal mode) begins with the formation of flat, adherent protrusions driving directional motility. Conversely, amoeboid migration depends on Rho/Rock‐dependent actomyosin contractility, driving blebbing‐like movements of loosely adherent cells. Conversion between these migratory modes, amoeboid‐to‐mesenchymal (AMT) and mesenchymal‐to‐amoeboid (MAT) transition, confers plasticity to metastatic tumor cell migration (Wolf and Friedl, 2006). The endocytic/recycling of Rac and α5β1 may control the migration into 3D matrices by promoting a mesenchymal mode of migration. GFP‐expressing, HeLa cells embedded in matrigel and treated (top) or not (bottom) with the Rock inhibitor Y27632 or expressing Rab5 (not shown) undergo an amoeboid‐to‐mesenchymal‐like cell shape transition in 3D. [Images are reprinted from Palamidessi et al. (2008) Copyright (2008), with permission from Elsevier].

Mechanistically, one important question raised by these findings concerns how signaling molecules are recycled to specific regions of the plasma membrane (as opposed to the bulk plasma membrane), to execute spatially restricted signaling. In the case of Rac, one possible answer may come from recent studies connecting localized Rac activation with integrin‐mediated adhesion and lipid raft internalization. These studies suggested that Rac positioning at and trafficking from and to plasma membrane specific locations may be regulated through raft endocytosis, needed, in turn, to specify its localized activity for the execution of relevant biological processes (2004, 2005, 1999, 2004). Thus, upon activation of integrin, sites of high Rac affinity become available on the plasma membrane preventing its internalization, which occurs following cell detachment in a Dynamin‐ and Caveolin‐dependent manner (del Pozo et al., 2005). Notably disruption of Dynamin function also results in segregation of active Rac in aberrant membrane invaginations, away from the plasma membrane, preventing the formation of regularly shaped lamellipodia, likely by blocking macropinocytic Rac internalization (Schlunck et al., 2004). It is thus tempting to speculate that Rac, activated through a process that requires CME (Palamidessi et al., 2008), is then delivered to specific regions of the plasma membrane, represented by lipid rafts (2004, 2005) whose non‐Clathrin internalization is prevented by integrin signaling. Within this context, Arf6‐dependent recycling appears to be the critical route controlling not only the redelivery of Rac (Palamidessi et al., 2008) and integrins (2004, 2005), but also of lipid raft, back to the plasma membrane, ultimately coordinating Rac signaling and directional migration with adhesion‐dependent cell growth (Balasubramanian et al., 2007).

The precise molecular determinants linking and regulating the activities of all these proteins are not entirely clear. One clue in this direction is provided by the observations that Arf6‐polarized recycling of cholesterol‐rich membranes requires microtubule and motors for directed delivery (Balasubramanian et al., 2007), suggesting integration between membrane trafficking and microtubules‐based transport. Additionally, a key signaling node of these pathways might be a large multimolecular unit composed of GIT1, a member of a family of GTPase‐activating proteins for ARF6, paxillin, a focal adhesion protein, and a complex including the Rac/Cdc42 exchanging factors PIX/Cool and the kinase PAK, which assembles all the critical elements controlling vesicle recycling, focal adhesion turnover, Rac and Cdc42‐directional migration (de Curtis, 2001).

The importance of spatial restriction of Rac signaling in controlling migratory protrusions is underscored by the discovery that the formation of CDR correlates with the ability of cells to acquire a mode of motility which is typical of mesenchymal moving cell, and to migrate in 3D (Suetsugu et al., 2003), properties that are defining features of metastatic invading cancer cells (Figure 3B). Recent experimental evidence mainly based on the use of intravital two‐photon analysis of invading cancer has revealed that individual tumor cells detaching from the original mass and adventuring into the surrounding tissue can adopt diverse modes of migration whose molecular determinants are only just starting to be elucidated [reviewed in Wolf and Friedl (2006)]. Amoeboid motility is characterized by the formation of uniformly distributed fast‐extending and fast‐retracting blebs, and relies on efficient actomyosin contractility (Wolf and Friedl, 2006). This type of motility is conserved throughout evolution and has been shown to have physiological relevance at the organismal level, such as during primordial cell migration in Zebrafish (Blaser et al., 2006). Mesenchymal‐type migration, conversely, relies on the formation of persistent and extended Rac‐dependent lamellipodia protrusions, and on the activation of metalloproteases (Wolf and Friedl, 2006). Remarkably, tumor cells can plastically switch from one mode of motility to the other dramatically increasing their chance to invade regardless of the variety, but rather taking advantage of the diversity of the microenvironmental conditions.

Rac signaling has been shown to be critical for mesenchymal migration (Sanz‐Moreno et al., 2008), suggesting that endocytic/recycling by ensuring Rac spatial restriction and polarization may also control the amoeboid‐to‐mesenchymal transition. Indeed, the expression of Rab5 in cultured melanoma cells, a model for amoeboid motility, induced a change from amoeboid to a more mesenchymal‐like morphology and movement. Conversely, inhibition of Rab5 in cultured colon carcinoma cells, a model for mesenchymal motility, resulted in a switch to a more amoeboid‐like morphology and motility (Palamidessi et al., 2008). Most intriguingly, Rab5 was shown to play a role also in normal cell migration because it is required in vivo in the guidance of primordial germ cell migration during zebrafish development (Palamidessi et al., 2008), suggesting that spatially‐regulated membrane trafficking, and trafficking of key signaling molecules, such as Rac, is an evolutionary conserved mechanism to orchestrate localized actin dynamics, polarized protrusive activity, directed cell motility, ultimately affecting the modes of motility of normal and cancer cells.

4.6. Membrane trafficking of adhesion receptors in the regulation of migratory and invasive programs

Cell migration proceeds by cycles of edge protrusion, adhesion, and retraction whose precise coordination in space and time is critical to promote cell locomotion. An optimal organization of actin filaments, myosin, and adhesion sites is, thus, required for fast migration, indicating that actin‐filament assembly, force generation, and adhesion are interdependent functions (Giannone et al., 2007).

Integrins, which are the major cell surface adhesion receptors for ligands in the extracellular matrix (Hynes, 2002), play a critical role in regulating cell migration. Integrins are heterodimeric, transmembrane proteins consisting of an α‐ and a ß‐chain and are involved in the transmission and interpretation of signals from the extracellular environment into various signaling cascades (Hynes, 2002). Several different mechanisms, including expression and subunit heterodimerization patterns, clustering and lateral diffusion in the plane of the plasma membrane, and interaction with the actin cytoskeleton and the inside of cells control their activity (Calderwood, 2004). Additionally, many integrins are continually internalized from the plasma membrane into endosomal compartments and are subsequently recycled (Caswell and Norman, 2006), prompting the proposal that the endo/exocytic cycle of these adhesion receptors is essential to control various aspects of cell locomotion. A number of excellent reviews have recently covered the molecular details and mechanisms through which various integrins by being internalized either through CME and NCE are subsequently recycled through Rabs ‐or Arf6‐dependent routes toward the cell front, thus maintaining a spatially‐restricted, polarized distribution of cycling extracellular matrix receptors within the confine locale of an advancing cell protrusion (Caswell and Norman, 2006). Remarkably, the integrin endo/exocytic cycle, similar to what has been proposed for motogenic receptors of the RTK and GPCR family, appears to be critical to promote directional migration in 2D and was recently shown to mediate the motility behavior of cells in 3D, ultimately controlling cell invasion. Clear examples of this kind are the studies on αVβ6, whose role in promoting haptotaxis and invasive migration in oral carcinoma is firmly established (Thomas et al., 2006). Disruption of αVβ6 CME‐internalization prevented haptotactic cell migration toward αVβ6 ligands and invasion through matrigel in organotypic invasion assays, which more closely model the tumor stroma (Ramsay et al., 2007). This provided the first demonstration that endocytic trafficking of adhesion receptor directly contributes to invasive migration, similar to what is observed during tumor metastatization.

Evidence, instead, that recycling routes control the ability of integrin‐mediated migration into fibronectin‐containing 3D‐extracellular matrix has been provided for α5β1, that associates with the small GTPase Rab25, a member of the Rab11 family, involved in vesicle recycling (Caswell et al., 2007) (Figure 3). Rab25 was shown to promote the formation of long pseudopodial extensions as cells migrate in 3D contexts, and both Rab25 and α5β1 integrin co‐localize in relatively static intracellular vesicular compartments within the distal tips of pseudopods (the lamellipodia equivalent in a 3D setting) (Caswell et al., 2007). Using a photoactivatable GFP‐α5 unit to analyze the dynamics of integrin trafficking in real‐time permitted to reveal that α5β1 integrin traffics bidirectionally between intracellular Rab25 vesicles and the plasma membrane within the confines of the pseudopodial tips. Additionally, Rab25 promoted the compartmentalization of a spatially restricted subpopulation of cycling α5β1 within the tip regions of extending pseudopods as cancer cells move forward through a 3D matrix (Caswell et al., 2007).

The scenario that is emerging from all these studies is that there is a “rush‐hour‐like” intense trafficking of vesicles and cargo molecules as a cell becomes motile, whose precise regulation is required to ensure polarized migratory response both in 2D and 3D settings. Mechanisms that may enable to guide different cargos to the right intinerary among this intricate road network are to either utilize the same route as bystander passengers or to coordinate their navigation so as to reach the same destination in a temporal and spatially controlled fashion. Some evidence in support of the latter notion is beginning to be provided. For instance, it has been shown that the there is a complex level of signal and trafficking crosstalk between αVβ3 and α5β1 integrins and the EGFR. Blockade of the adhesive function of αVβ3 promoted Rabs‐dependent polarized recycling of α5β1 leading to the acquisition of rapid/random movement on two‐dimensional substrates and to a marked increase in fibronectin‐dependent migration of tumor cells into three‐dimensional matrices. Remarkably, the enhance recyling of α5β1 had no effect on its adhesive properties, rather it facilitated the association of EGFR with the recycling machinery promoting the coordinated recycling of these two receptors, increasing EGFR autophosphorylation and activation of the pro‐invasive kinase AKT (Caswell et al., 2008). How general is co‐trafficking of diverse receptors remains to be investigated, nevertheless the possibility that co‐navigation not only of membrane‐bound receptors, but also of some of the key signaling effectors may have evolved as an effective way to achieve polarization will surely receive increasing attention.

4.7. Regulating epithelial polarity by directional trafficking

To form functional and organized tissues, cells need to control their morphology, especially during certain development stages. Epithelial cells can, for example, lose their attachment to each other, depolarize and undergo a process called epithelial–mesenchymal transition (EMT) (Thiery, 2003). At other stages during development, epithelia change their shape as a result of a coordinated rearrangement of individual epithelial cells. Recent experiments have demonstrated a crucial role for endocytosis and recycling of cell adhesion molecules during each of these developmental processes.

Epithelial cells are characterized by a subdivision of their plasma membrane into an apical and a baso‐lateral domain. Cell–cell junctions at their boundary block diffusion between the apical and basolateral domains and also serve as adhesion sites to ensure mechanical connections between cells. Key components of epithelial junctions in vertebrates are the transmembrane protein E‐cadherin, which define the so‐called adherens junction, and the Zonula Occludens‐1 (ZO‐1), which was the first protein to be identified in tight junction. In this latter structure, three distinct types of integral‐membrane proteins have been identified: Occludin, Claudin and Jam [reviewed by Tsukita et al. (2008)]. The trafficking of tight junctional proteins has only recently begun to be analyzed. Conversely, E‐cadherin endocytic/recyling is well established and it is emerging as an important mechanism in assembling and disassembling epithelial tissues (Yap et al., 2007).

Like other transmembrane proteins, Cadherins are synthesized in the ER, modified in the Golgi, and trafficked to the baso‐lateral cell surface. Once at AJs, cadherins can be endocytosed into Rab5‐positive early endosomes, and either recycled back to AJs via Rab11‐positive recycling endosomes, or sent via the Rab7‐ and Hrs‐positive MVB to lysosomes for destruction (Bryant and Stow, 2004).

While E‐cadherin degradation is required for EMT, it has to be prevented in functional epithelia to maintain apical–basal polarity. Surprisingly, E‐cadherin localization to adherens junctions is not maintained by preventing its endocytosis, but by redirecting the protein to the plasma membrane through a recycling pathway. E‐cadherin recycling has been observed both in mammalian cells (Paterson et al., 2003) and in Drosophila (Langevin et al., 2005). It seems to be regulated by the small GTPase Rab11 and by the exocyst (Wu et al., 2005) – a multiprotein complex that acts as a Rab11 effector and mediates polarized membrane delivery (Hsu et al., 2004). At least in Drosophila, exocyst activity is crucial to deliver E‐cadherin from the recycling endosomal compartment to the plasma membrane (Beronja et al., 2005) or specifically to epithelial cell junctions (Blankenship et al., 2007; Langevin et al., 2005). Regulating recycling of junctional proteins can modulate morphogenesis; for example, Rab11 function is turned down in some tracheal cells, to reduce cell surface cadherin and allow cell rearrangement (Shaye et al., 2008). In contrast to trafficking into the degradative pathway, which requires the activation of RTKs and subsequent ubiquitination of E‐Cadherin, recycling seems to be the default pathway of endocytosed E‐cadherin (Paterson et al., 2003). Thus, this constant turnover of junctional material seems to allow epithelial cells to rapidly adjust polarity and cell–cell contacts in response to extracellular stimuli. In keeping with the notion that there is a tight coupling between E‐cadherin trafficking and the establishment of cell polarity, in Drosophila, it has been shown that the protocadherin Flamingo can recruits the exocyst component Sec5 to particular areas of the plasma membrane to direct the delivery of E‐cadherin (Classen et al., 2005). Since Flamingo is part of the well‐characterized planar cell polarity pathway (Strutt and Strutt, 2005), these findings are an instructive example of how such genes can act directly on the basic trafficking machinery of a cell to influence the architectural organization of an organ.

The molecular components underlying the connection between E‐cadherin trafficking and cell polarity have recently been enriched by the demonstration that the small GTPases Cdc42 and its effector, the Par3/Pa6 polarity complex, are critical for stabilizing dynamics, basolateral, adherens junction through regulation of apical endocytosis in the Drosophila neuroectodermal epithelium (Harris and Tepass, 2008) and in the notum or dorsal thorax (Georgiou et al., 2008; Leibfried et al., 2008). Cdc42 appears to act at multiple steps of the trafficking of junctional proteins. Its loss caused an increase in the endocytotic uptake of apical proteins, including apical polarity factors such as Crumbs, which are required for AJ stability. The Par3/Par6 polarity complex is involved in this latter process, corroborating the first demonstration of the role this protein assembly in endocytosis of vitellogenin in C. elegans oocytes and coelomocytes (Balklava et al., 2007). Similarly, the Cdc42 effectors WASP (Wiskott–Aldrich syndrome protein) and components of the Arp2/3 complex caused a defective E‐cadherin internalization (Georgiou et al., 2008; Leibfried et al., 2008), suggesting the participation of actin dynamics also in the endocytic maintenance of adherens junction stability, albeit the precise mechanisms through which Cdc42 may concomitantly control the Par complex and WASP/Arp2/3 remain elusive. In addition, Cdc42 has a second function in regulating trafficking since it is required for the progression of apical cargo from the early to the late endosome. One model proposed to account for all these observations is that Cdc42 may slow the entry of apical proteins into the endocytic pathway, while accelerating protein traffic from early to late endosome. In its absence apical protein are rapidly internalized, but become trapped in enlarged Hrs‐positive endosomes rather than being sent to the lysosome for degradation. An alternative explanation is that Cdc42 promotes recycling of proteins back to the cell surface, consistent with the localization of dominant‐negative Cdc42 to the enlarged endosome. In remodelling tissue, AJs must be inactivated to allow cells to change partners, but epithelial integrity must not be lost during remodelling. AJ proteins may treadmill, constantly being endocytosed and then recycled back to the plasma membrane, allowing cells to slide in relationship to one another. In Cdc42 mutants, reduced apical protein recycling would disrupt epithelial organization, with tissues undergoing extensive remodelling most sensitive to this reduction. Ultimately, cell fate cues must control protein traffic differentially, and the identification of how protein traffic is regulated in developing tissues is the next major challenge.

5. Perspective

Our view of the functional implications of endocytosis and recycling has significantly changed in the course of the last decade. There are nowadays compelling biochemical and genetic evidence that support the notion that intracellular trafficking is critical to resolve spatial information essential for the control of polarized and asymmetric cell processes. Our understanding, however, of the molecular and regulatory circuitry guiding the trafficking of cargos and membrane through the intracellular road network is still in its infancy. As the molecular players come into focus, it will be possible to build dynamic maps integrating signaling cascades and trafficking organelles and routes. This will be critical to unravel the mechanisms controlling physiological and pathological processes that must respond to spatial cues, such as tissues morphogenesis or cell migration, or subvert spatial confinements, such as during tumor invasion.

This latter process is much more complex than initially anticipated. Cancer cells possess a broad spectrum of migration and invasion mechanisms. These include both individual and collective cell‐migration strategies (Wolf and Friedl, 2006). Importantly, metastatic, invasive tumor cells can switch between these various modes of motility, plastically adapting their invasive programs to the different microenvironmental conditions. These adaptive responses provide migratory ‘escape’ strategies after pharmacotherapeutic intervention, by prompting alternative mechanisms of cancer cell dissemination in tissues that overcome single‐pathway‐hitting pharmacological weapons (Jechlinger et al., 2002; Wolf and Friedl, 2006). As a case in point, pharmacological inhibition of metalloproteases (which targets primarily mesenchymal motility) had been proposed as a tool to inhibit tumor spreading and dissemination, with encouraging results in animal model system (Drummond et al., 1999). However, clinical trials in several types of human cancers have yielded disappointing results (Arvelo and Cotte, 2006). The ability of tumor cells, for instance, to convert from metalloprotease‐dependent (mesenchymal) to protease‐independent (amoeboid) motility might account, at least in part for these failures. In this framework a detailed molecular understanding of the “switch” is essential to improve our ability to design effective anti‐invasive and anti‐metastatic therapies, by preventing cancer cells from interconverting between migratory modes. Endocytosis and recycling are emerging as critical processes for controlling this switch. Indeed, spatially‐regulated membrane trafficking, and trafficking of key signaling molecules, such as Rac, or adhesion receptors, such as integrins, control the ability of cells to move in a mesenchymal‐like fashion, suggesting that endocytosis and recycling cycles are implicated in the plasticity of tumor cell migration and invasion. As a consequence, multiple components of endocytosis/recycling could potentially become the targets of anti‐invasive strategies; this idea certainly warrants further investigations.

One additional aspect that has received little attention is the relationship between transcriptional regulation and membrane trafficking. An unexpected breakthrough in this direction has been recently reported (Wang et al., 2009). In this study, the activation of hypoxia‐inducible factor (HIF1), a molecule central to oxygen sensing, was shown to repress transcriptionally Rabaptin‐5, a critical Rab5‐effector, blocking Rab5‐mediated endosome fusion. This resulted in inhibition of EGFR internalization and degradation, and enhancement of EGFR‐dependent signaling, enabling cells to proliferate and survive in the unwelcoming environment of a hypoxic tumor mass. This study established a new link between hypoxia‐dependent transcriptional regulation and endocytic control of cell signaling, predicting unexpected levels of interconnection between microenvironment, gene expression and intracellular trafficking and signaling. The exploration of this uncharted territory will surely reserve future surprises and likely provide new tools and strategies of therapeutic intervention.

Disanza Andrea, Frittoli Emanuela, Palamidessi Andrea, Scita Giorgio, (2009), Endocytosis and spatial restriction of cell signaling, Molecular Oncology, 3, doi: 10.1016/j.molonc.2009.05.008.