Integrin trafficking at a glance

Integrin endocytosis has been studied extensively in the context of viral entry (Pellinen and Ivaska, 2006), but mechanisms that regulate integrin internalisation in the absence of pathogens have been identified. Integrins can be internalised through the major endocytic routes (Box 1), including the formation of circular dorsal ruffles (CDRs) during macropinocytosis (Gu et al., 2011), and clathrin-dependent and -independent endocytic pathways. Direct recruitment of endocytic regulators to integrins constitutes a general mechanism that can drive integrin endocytosis. Examples of this include the recruitment of HS1 associated protein X-1 (HAX-1) to the cyoplasmic tail of integrin β6 (Ramsay et al., 2007) and the recruitment of Rab21 to the cytoplasmic tail of integrin α subunits (Pellinen et al., 2006). Protein kinase C alpha (PKCα) controls caveolar endocytosis (Smart et al., 1995), and can itself bind to β1 integrins (i.e. integrins that contain a β1 subunit and any α subunit) and regulate integrin internalisation (Ng et al., 1999). In endothelial cells, recruitment of active α5β1 integrin to the neuropilin–GIPC1 (for GAIP C-terminus-interacting protein) complex regulates endocytosis of this integrin directly from fibrillar adhesions (Valdembri et al., 2009). Xenopus GIPC1 (also known as kermit2) regulates α5β1 integrin internalisation into Rab21-positive vesicles (Spicer et al., 2010), which suggests that mechanisms that regulate integrin internalisation are conserved.

The clathrin adaptors disabled homologue 2 (DAB2) and NUMB interact, through their phosphotyrosine-binding (PTB) domains, with the NPxY-motifs in the β-integrin cytoplasmic tails (Calderwood et al., 2003), and several lines of evidence support the notion that clathrin-adaptors regulate integrin endocytosis. Microtubule-induced focal adhesion (FA) disassembly has been shown to require clathrin-dependent endocytosis (CDE) (Chao and Kunz, 2009; Ezratty et al., 2009). Indeed clathrin and DAB2 localise to FAs in the mid-region of moving cells and regulate α5β1 integrin internalisation to facilitate migration. NUMB binds to β1 and β3 integrins, and localises around focal contacts at the leading edge of migrating cells together with clathrin. Phosphorylation of NUMB downstream of the PAR3 (also known as PARD3)–atypical PKC (aPKC) complex regulates integrin internalisation at these sites to promote migration (Nishimura and Kaibuchi, 2007). Inactive integrins are also internalised in a clathrin-dependent manner: DAB2 and AP2 regulate endocytosis of α1β1, α2β1 and α3β1 (but not α5β1) integrins from the dorsal surface of the cell, and the rate of DAP2–AP2-mediated endocytosis correlates positively with migration specifically towards ligands for these integrins (Teckchandani et al., 2009).

Accumulating evidence indicates that integrins can follow clathrin-independent endocytic routes. Indeed, β1 integrins are found within clathrin-independent carriers (CLICs, see Box 1) (Howes et al., 2010a). Rab21 overexpression circumvents the requirement for CDE to internalise β1 integrins in dividing cells by driving a form of clathrin-independent endocytosis (CIE) (Pellinen et al., 2008). Cholesterol depletion experiments have shown that integrins can enter the cell in a raft-dependent fashion (Fabbri et al., 2005), and α5β1 and αvβ3 integrins are thought to associate with caveolin-1 (Gálvez et al., 2004; Wickström et al., 2002). More direct evidence supports a role for caveolae in the internalisation of α5β1 in myofibroblasts (Shi and Sottile, 2008), α2β1 in an osteosarcoma cell line (Upla et al., 2004) and active β1 integrins in bone marrow mesenchymal stem cells (BMMSCs; Du et al., 2011).

Although integrin endocytosis can be constitutive and ligand-independent, cell adhesion to the ECM through other adhesion receptors can promote integrin endocytosis. For instance, myelin-associated glycoprotein (MAG) inhibits axonal growth in neurons and repels growth cones by increasing the local Ca²⁺ concentration, thereby stimulating clathrin-dependent endocytosis of β1 integrins to promote growth cone turning (Hines et al., 2010). Syndecan-4 can act as a co-receptor for fibronectin (alongside integrins). In addition, syndecan-4 engagement controls the availability of α5β1 at the plasma membrane in fibroblasts through a mechanism that is consistent with caveolar endocytosis of the integrin (Bass et al., 2011).

Internalised integrins are rapidly trafficked to early endosomes (EEs), where cargoes are sorted for degradation or recycling (Box 2; Caswell and Norman, 2006). In the presence of fibronectin ligand, α5β1 integrin can be routed to late endosomes and degraded (Dozynkiewicz et al., 2012; Lobert et al., 2010; Tiwari et al., 2011), but the majority of internalised integrins are rapidly recycled back to the plasma membrane (Bretscher, 1989; Bretscher, 1992). Integrins can follow the major recycling routes of other cargoes, such as the transferrin receptor (TfnR, also known as TFRC) (Box 2) (Caswell and Norman, 2006; Caswell et al., 2009; Pellinen and Ivaska, 2006). However, although there are shared mechanisms between these two functionally distinct cargoes, there are also integrin-specific elements that provide spatial and temporal control of adhesion receptor availability at the plasma membrane.

αvβ3 integrins recycle through the Rab4-dependent ‘short-loop’ pathway (Box 2), which returns the heterodimers from EEs back to the plasma membrane without transiting through the perinuclear recycling compartment (PNRC) (Roberts et al., 2001; Jones et al., 2009; di Blasio et al., 2010). PKD1 (for protein kinase D1), when autophosphorylated on Ser916 in response to growth factor stimulation, interacts with the extreme C-terminus of integrin β3, and this interaction is required for Rab4-dependent recycling of this integrin (Woods et al., 2004; White et al., 2007; di Blasio et al., 2010). Rab4-dependent trafficking of αvβ3 is key to the control of directionally persistent migration (di Blasio et al., 2010; White et al., 2007; Woods et al., 2004) and branching morphogenesis of endothelial vessels (Jones et al., 2009). Although α5β1 does not follow the Rab4 recycling pathway in platelet-derived growth factor (PDGF)-stimulated fibroblasts (Roberts et al., 2001), recent studies have indicated that β1 integrins can follow a Rab4-dependent route to the plasma membrane: epidermal growth factor (EGF) promotes β1 integrin recycling in HeLa cells in a manner that requires the raft-associated membrane protein supervillin (Fang et al., 2010), and recycling of inactive β1 integrins is Rab4-dependent in breast cancer cell lines (Arjonen et al., 2012).

Integrins can also follow a second, ‘long-loop’ recycling pathway through the PNRC (Box 2), and this is dependent on the GTPases Rab11 and Arf6 and is linked to cell migration (Powelka et al., 2004; Caswell and Norman, 2006; Roberts et al., 2004). For β1 integrins, many of the mechanistic details are shared with other cargoes. EHD1 (for EH-domain containing 1), which is recruited to tubules emanating from the PNRC by MICAL-like 1 (MICALL1) and phosphoinositides, regulates the recycling of both β1 integrins and the TfnR (Jović et al., 2007; Jović et al., 2009; Naslavsky and Caplan, 2011; Sharma et al., 2009). In addition, Myotubularin, a phosphoinositide phosphatase, is required for the exit of integrins from intracellular compartments in Drosophila (Ribeiro et al., 2011). SNARE function is crucial in docking and fusion of vesicles throughout the endosomal system, and several Q-SNAREs (e.g. syntaxin 6, SNAP29, syntaxin 4 and SNAP23) and R-SNAREs (e.g. VAMP2 and VAMP3) have been shown to regulate β1 integrin trafficking (Veale et al., 2010; Tiwari et al., 2011; Skalski and Coppolino, 2005; Rapaport et al., 2010; Hasan and Hu, 2010).

Although integrins share much of the basic machinery for Rab11-dependent recycling with the TfnR, stages of the long-loop recycling pathway that are specifically required for integrin recycling have been identified. Rab21 binds to a conserved motif within the α-subunits of β1 integrin heterodimers and specifically promotes their internalisation and trafficking through early endosomes to the PNRC (Pellinen et al., 2008; Pellinen et al., 2006). Here, p120RasGAP (also known as RASA1) displaces Rab21, thereby promoting the release of the integrin-containing vesicles from the perinuclear region and their recycling to the plasma membrane (Mai et al., 2011). In carcinoma cells, Rab-coupling protein (RCP), a member of the Rab11-family interacting proteins (RAB11FIPs), which act as Rab11 effectors, has a crucial role in regulating integrin recycling but is dispensable for TfnR recycling through Rab11 recycling endosomes (Caswell et al., 2008; Muller et al., 2009). Both α5β1 and αvβ3 integrins follow a Rab11-dependent route in unstimulated fibroblasts. This requires the inactivation of glycogen synthase kinase 3 beta (GSK3β) through phosphorylation by AKT (also known as PKB) (Roberts et al., 2004). Serum stimulation can also promote trafficking of β1 integrins through this route, and AKT activity is required to phosphorylate ACAP1 (for ArfGAP with coiled-coil, ankyrin repeat and PH domains 1), which recruits β1 integrins to a recycling coat complex that contains clathrin (Powelka et al., 2004; Li et al., 2005; Li et al., 2007). PKCε phosphorylates vimentin, a type III intermediate filament protein, and releases β1-integrin-containing vesicles from intermediate filaments in the perinuclear region to allow recycling of β1 integrins (Ivaska et al., 2005; Ivaska et al., 2002). The events regulated by AKT and PKCε are specific to integrin trafficking, and are not required for recycling of TfnR.

Many of the studies described above provide compelling evidence that integrin trafficking contributes directly to cellular processes such as cell migration. Endocytic mechanisms can contribute to the turnover of adhesion complexes, and, in this way, directly regulate interactions with the ECM and proteins involved in adhesion signalling [e.g. focal adhesion kinase (FAK) phosphorylation (Ezratty et al., 2005; Ezratty et al., 2009)]. It is unclear, however, whether trafficking pathways can influence the activation status of integrins or the ability of integrins to mediate adhesive contacts with the substratum, and in some cases this is evidently not the case (Caswell et al., 2007; Caswell et al., 2008). Tracking integrin trafficking using antibody surface-labelling approaches has indicated that internalised integrins can subsequently localise to focal complexes and focal adhesions, but only after extended periods (i.e. hours) following antibody exposure (Gu et al., 2011). Given that the t_1/2 of integrin recycling is often <15 minutes, this could indicate that integrins do not recycle directly to focal adhesions (Caswell and Norman, 2006). Accumulating evidence suggests that instead integrin trafficking influences intracellular signalling, either directly or by influencing trafficking of other cargoes.

Integrin trafficking pathways can directly impact on Rho GTPase signalling. Rab21-mediated recycling of α5β1 is required for the activation of RhoA at the cleavage furrow and, hence, the completion of cytokinesis (Pellinen et al., 2008). Rho GTPases are master regulators of the cytoskeleton, and balancing RhoA and Rac activities is crucial for controlling the directional persistence of migrating cells (Danen et al., 2005). In fibroblasts, Rab4-dependent recycling of αvβ3, and the activity of this integrin, suppresses recycling of α5β1, thereby promoting directionally persistent migration that is characterised by a broad leading lamellipodium (Caswell et al., 2008; White et al., 2007). Inhibition of αvβ3 recycling or ligand binding enhances recycling of α5β1 and signalling through the ROCK (RhoA–Rho-associated, coiled-coil containing protein kinase) pathway, which, in turn, leads to phosphorylation of the downstream effector cofilin and rapid random migration (White et al., 2007).

Integrin trafficking influences the ability of cells to move in three-dimensional matrices, as well as on two-dimensional substrates. Hypoxia-driven invasion through a laminin-rich matrix requires Rab11, and there is evidence that this involves the Rab11-dependent recycling of α6β4 (Yoon, 2005). αvβ6 promotes invasion in squamous cell carcinoma, and the interaction of this integrin with HAX1 allows its clathrin-dependent endocytosis, which in turn promotes carcinoma cell motility in organotypic culture (Ramsay et al., 2007).

Inhibition of αvβ3, or expression of cancer-associated mutant forms of p53, promotes α5β1 recycling in carcinoma cell lines (Caswell et al., 2008; Muller et al., 2009). This requires the interaction of β1 integrin with RCP (Caswell et al., 2008), which is itself upregulated in breast cancer (Zhang et al., 2009). The actin nucleation promoting factor (NPF) WASH (for WAS protein family homologue) has a role in α5β1 recycling through this pathway, presumably by remodelling actin on endosomes (Zech et al., 2011). RCP-dependent α5β1 trafficking promotes invasive migration in three-dimensional matrices, which is characterised by the extension of invasive pseudopods. The production of phosphatidic acid within the pseudopod tip is required to localise RCP and permit integrin recycling (Rainero et al., 2012). Rather than influence α5β1 activity, α5β1 and RCP recruit EGF receptor 1 (EGFR1) and coordinate its recycling, which, in turn, potentiates EGFR activation and downstream signalling to promote invasion (Caswell et al., 2008; Muller et al., 2009). In this context, integrins and RCP act as key components of a ‘recyclosome’ complex, which recruits receptor tyrosine kinases (RTKs), including EGFR1, ErbB2 and Met, promotes their recycling and potentiates signalling to induce metastasis (Muller et al., 2009; Caswell et al., 2008; Muller et al., 2012; P. T. Caswell, unpublished observations). Interestingly, in neuronal axons, Rab11 and RCP can control trafficking of β1 integrins to promote axonal extension (Eva et al., 2010), indicating that this mechanism is not unique to cancer cells.

Overexpression of Rab25 (also known as Rab11c) is associated with aggressive ovarian cancers (Cheng et al., 2004). Rab25 specifically regulates trafficking of α5β1 by directly interacting with β1 integrin (Caswell et al., 2007). Inactive integrins that become internalised into Rab25 vesicles at the front of invasive cancer cells are recycled back to the plasma membrane within this subcellular region, thus promoting their spatial restriction (Caswell et al., 2007; Caswell and Norman, 2008; Caswell et al., 2009). Those integrins that remain in the active conformation following internalisation, however, are sorted through Rab25-positive late endosomes to lysosomes, moving from the front of the cell towards the rear. CLIC3 regulates the subsequent recycling of active α5β1 from lysosomes to the plasma membrane. This allows Rab25 to control pseudopod extension at the front of the cell and coordinate retraction of the cell body by trafficking active integrins towards the rear of the cell to promote integrin signalling and forward movement (Dozynkiewicz et al., 2012).

Integrin trafficking pathways can indirectly influence intracellular signalling by controlling endocytic trafficking of other cell surface receptors. In BMMSCs, soft substrates promote endocytosis of β1 integrins through caveolae, and this is required for the internalisation of bone marrow morphogenetic protein (BMP) receptor IA (BMPRIA). BMPRIA colocalises with integrins in an intracellular compartment, which implies that these endocytic cargoes co-internalise, and that this, in turn, inhibits BMP-induced SMAD signalling, which allows cells to differentiate along the neuronal rather than osteogenic or myogenic lineage (Du et al., 2011).

αvβ3 integrin controls receptor trafficking, suppressing the recycling of α5β1 integrin, and thereby leads to changes in Rho GTPase signalling in fibroblasts and promotes receptor-tyrosine kinase trafficking and signalling in carcinoma cells as discussed above. Inhibiting αvβ3 with small molecule inhibitors such as Cilengitide in endothelial cells drives Rab4-dependent vascular endothelial growth factor receptor 2 (VEGFR2) recycling, protecting this receptor from degradation in the presence of VEGF ligand and increasing its levels on the cell surface (Reynolds et al., 2009). This promotes VEGF-driven endothelial cell migration, sprouting of aortic explants and tumour angiogenesis in vivo, which ultimately leads to enhanced tumorigenesis.

Clearly, integrins are cargoes of intracellular trafficking pathways; however, the relationship between integrins and endocytic traffic is complex because integrins themselves govern signals that control endocytic flux. In some instances adhesion negatively regulates endocytosis: engagement of integrin β1 by ligand can slow the dynamics of clathrin-coated structures, and adhesion of cells to fibronectin decreases the rate of TfnR endocytosis (Batchelder and Yarar, 2010). By contrast, β1 integrin engagement promotes endocytosis of the platelet-derived growth factor receptor (PDGFR) during fibroblast chemotaxis. In this scenario, integrins are required to maintain the stability of the actin NPF neural Wiscott–Aldrich syndrome protein (N-WASP), which controls the formation of CDRs and the internalisation of PDGF through macropinocytosis (King et al., 2011). Signalling from β1 integrins is also required for the fusion of vesicle-associated membrane protein 7 (VAMP7) vesicles with the plasma membrane in extending neurites on laminin substrates (Gupton and Gertler, 2010).

Integrin engagement also regulates the trafficking of caveolae and/or lipid rafts. Loss of adhesion triggers internalisation of lipid rafts, which is controlled by the shift of phosphorylated caveolin from adhesion complexes to caveolae (del Pozo et al., 2005; del Pozo et al., 2004). Following re-adhesion, lipid rafts recycle to the plasma membrane from the PNRC in an Arf6-dependent manner (Balasubramanian et al., 2007). RalA regulates exocytosis through the exocyst complex (Balasubramanian et al., 2010), which provides a platform for the activation of Rac. Keratinocytes that lack integrin β1, or the adhesion signalling protein integrin-linked kinase (ILK), also show a defect in the membrane targeting of caveolae. Caveolar endocytosis occurs independently of β1 integrins and ILK, but integrin signalling promotes the recruitment of a complex containing IQGAP (IQ motif containing GTPase activating protein) and mDia1 (also known as DIAPH1) to the plasma membrane, which in turn locally captures and stabilises microtubules to permit re-exocytosis of caveolin-1 (Wickström et al., 2010).

Integrin trafficking is a dynamic process that is important for the regulation of cellular processes, including migration and cytokinesis. Technical limitations have hindered our understanding of integrin trafficking in vivo, but these are being overcome in some model systems (Yuan et al., 2010; Ribeiro et al., 2011), and furthering these studies to directly visualise trafficking events in vivo is a priority. Trafficking pathways can promote spatial restriction or en masse movement of integrins, and, interestingly, some regulators, such as Rab25, have a role in both (Caswell et al., 2007; Dozynkiewicz et al., 2012). Further studies are necessary to determine the sorting steps that control the decision to recycle integrins to the membrane from which they were initially internalised or to more distal sites. It will also be important to determine how the directional movement of integrins interfaces with downstream signalling pathways.

Unlike most cargoes, integrins are almost uniquely positioned in that they are cargoes for pathways that they themselves regulate. Further investigation is needed to determine whether integrins form central components of multi-cargo trafficking complexes that are controlled by adhesion signalling. This also raises an interesting question: are integrins able to elicit signals when they are transiting the endosomal system? The presence of active integrins on endomembrane compartments has now been noted by several investigators, and the recent identification of the integrin inactivator sharpin (Rantala et al., 2011) suggests that integrins could remain active on intracellular compartments until inactivated. Endomembranes are surfaces that provide a scaffold for signalling pathways such as the AKT (Schenck et al., 2008) and ROCK–myosin light chain 2 (MLC2) pathways (Sturge et al., 2006). This opens up the exciting possibility that integrins could associate with downstream signalling components on endosomes to propagate signals from a lipid and protein environment distinct from the adhesion plaque.