EHD1 regulates β1 integrin endosomal transport: effects on focal adhesions, cell spreading and migration

Integrins are the major eukaryote receptors for cell adhesion to the extracellular matrix (ECM). Upon ligand binding, heterodimeric α/β integrin receptors cluster to cytoskeleton-linked focal adhesion complexes and induce the recruitment of kinases and adaptor proteins leading to the transduction of signals to the cell interior (reviewed by Hynes, 2002). The binding of β1 integrins to the ECM, including laminin, collagen, vitronectin (VN) and fibronectin (FN), stimulates a variety of cellular signaling pathways that may lead to gene expression, proliferation, apoptosis, cell survival, cell spreading, migration, invasion, metastasis and angiogenesis (reviewed by Varner and Cheresh, 1996). Integrins can also transduce inside-out signals, whereby intracellular signaling influences the affinity of integrins for their ECM ligands (Hynes, 2002).

To function and transmit signals, β1 integrins must localize to the plasma membrane and be able to relocate to focal adhesions upon ligand binding. Accordingly, the internalization process is crucial in regulating the levels of integrins localized to the plasma membrane and in maintaining functionality of integrin-mediated downstream effects. Whereas some integrins contain a cytosolic NITY motif conforming to the classical NxxY internalization motif found on receptors internalized through AP-2/clathrin-coated pits, most integrins (including β1 integrin) appear to be internalized in a clathrin-independent manner (Altankov and Grinnell, 1995; Ylanne et al., 1995) that depends on the small GTPase Arf6 (Brown et al., 2001). Indeed, integrin internalization is regulated by protein kinase C-α (Ng et al., 1999) and dynamin (Ezratty et al., 2005; Ng et al., 1999), and several studies provide evidence that this process occurs either through caveolae and/or through detergent-resistant lipid microdomains (Upla et al., 2004; Wary et al., 1996; Wary et al., 1998).

Once internalized, integrin receptors return to the plasma membrane by a recycling route through the endocytic pathway (reviewed by Caswell and Norman, 2006; Jones et al., 2006; Pellinen and Ivaska, 2006). Certain integrin receptors, such as αvβ3, are recycled rapidly to the plasma membrane through a short loop, whereas α5β1 returns to the cell surface through a transitory recycling compartment (long-loop) (Roberts et al., 2001). Therefore, influencing the rate of integrin receptor recycling through these pathways is another crucial way in which integrin function may be modulated, and it has been widely accepted that the recycling of internalized integrins to the leading lamellae of a migrating cell affects its motility (Bretscher, 1996). Recent studies demonstrate that protein kinase C (Becker and Hannun, 2003; Ivaska et al., 2005; Ng et al., 1999) and protein kinase D1 (Woods et al., 2004) play important roles in controlling β1 integrin recycling to the plasma membrane. In addition, the recycling of integrins depends upon many of the same endocytic regulatory proteins that control the transport of other internalized receptors. In particular, the GTP-binding proteins Rab4 (Roberts et al., 2001), Rab11 (Powelka et al., 2004) and Arf6 (Powelka et al., 2004), and recently Rab21 and Rab5 (Pellinen et al., 2006), have been implicated in integrin recycling.

Based on these studies, it is likely that other regulators of endocytic recycling and proteins that physically and/or functionally interact with Rab-family proteins and Arf6 play a role in the regulation of integrin trafficking. One family of proteins that fits this description is the C-terminal EHD protein family (reviewed by Miliaras and Wendland, 2004; Montesinos et al., 2005; Naslavsky and Caplan, 2005), which has been implicated in endocytic transport and recycling of a variety of receptors (Caplan et al., 2002; Guilherme et al., 2004; Lin et al., 2001; Park et al., 2004; Picciano et al., 2003; Rotem-Yehudar et al., 2001; Shao et al., 2002). In mammals the four highly homologous paralogs (EHD1, EHD2, EHD3 and EHD4) all contain nucleotide-binding motifs (Lee et al., 2005; Naslavsky et al., 2006), a central coiled-coil involved in homo- and/or hetero-oligomerization (Galperin et al., 2002; Lee et al., 2005; Naslavsky et al., 2006), and a conserved C-terminal Eps15-homology (EH)-domain that binds to proteins containing the tripeptide asparagine-proline-phenylalanine (NPF) (Braun et al., 2005; Guilherme et al., 2004; Naslavsky et al., 2004; Naslavsky et al., 2006; Rotem-Yehudar et al., 2001; Xu et al., 2004).

Recent studies have shown that EHD proteins functionally interact with Arf6 and coordinate recycling events with Rab4 and Rab11. For example, EHD1 colocalizes with and is regulated by the guanine nucleotide-binding status of Arf6 (Caplan et al., 2002). Moreover, EHD4 binds to the cell fate determinant, Numb, whose subcellular localization also depends upon Arf6 nucleotide binding (Smith et al., 2004). In addition, recent studies demonstrate a direct interaction between EHD proteins and effectors of Rab4 and Rab11. The divalent Rab4/Rab5 effector Rabenosyn-5 interacts with EHD1 and its paralogs and plays a role in transport from early endosomes to the endocytic recycling compartment (Naslavsky et al., 2004). More recently, an interaction between the Rab11-effector Rab11-FIP2 and both EHD1 and EHD3 was identified, highlighting a coordinated recycling mechanism jointly controlled by Rab11 and EHD proteins (Naslavsky et al., 2006). Given the involvement of Rab-family proteins and Arf6 in integrin-mediated transport and recycling, we now demonstrate that EHD1 plays an essential role in regulating the trafficking of β1 integrins and their downstream functions.

To assess whether EHD1 affects the subcellular itinerary of β1 integrins, we used an RNAi approach in HeLa cells, effectively reducing EHD1 levels by 80-90% (Fig. 1A) (Naslavsky et al., 2004). To follow β1 integrin trafficking, we used a modified version of the assay developed by Powelka et al. using the 12G10 anti-β1 integrin antibody to label surface β1 integrins (Powelka et al., 2004). Cells were then allowed to internalize integrin-antibody complexes at 37°C for 2 hours (pulse), followed by a 2-hour incubation in the absence of antibody (chase) to monitor the distribution of internalized β1 integrins over time. Internal β1 integrins in mock-treated cells were mainly dispersed in peripheral vesicles (Fig. 1C). However, cells treated with EHD1-RNAi displayed an accumulation of β1 integrin-containing vesicles in the perinuclear region (Fig. 1D). Quantification of the level of accumulated internal β1 integrin showed a greater than twofold increase in EHD1-RNAi-treated cells (Fig. 1E), suggesting a delay in integrin recycling and accumulation at the endocytic recycling compartment (ERC). Since it has been demonstrated that certain β1 integrin heterodimers recycle to the plasma membrane (reviewed by Bretscher, 1992; Caswell and Norman, 2006), we hypothesized that loss of EHD1 might decrease the rate of β1 integrin recycling to the plasma membrane. To this aim, mock- and EHD1-RNAi-treated HeLa cells were subjected to flow cytometry-based recycling assays to compare levels of β1 integrins that have returned to the plasma membrane. Non-internalized anti-β1 integrin antibodies were stripped off the plasma membrane after the pulse by a brief acid rinse, a treatment that has been commonly used to follow trafficking of endogenous and chimeric proteins without interfering with the kinetics of their itinerary (Hemar et al., 1995). After a 2-hour chase at 37°C, antibody-integrin complexes that recycled back to the plasma membrane were detected with secondary antibodies under non-permeabilizing conditions. As shown (Fig. 1B), reduction of EHD1 expression typically decreased the levels of recycled β1 integrins localized to the plasma membrane by ∼30%. To assess whether β1 integrins traffick through EHD1-containing membranes, we pulsed GFP-EHD1-transfected HeLa cells with antibodies against β1 integrins (Fig. 1F). As demonstrated in the representative 0.4 μm z-section, the EHD1 structures tend to extend radially from the perinuclear recycling compartment region, and it was possible to distinguish individual vesicular membranes in three-dimensions containing both EHD1 and β1 integrins (see crosshairs and arrows), suggesting that β1 integrins indeed traverse an EHD1-regulated recycling pathway.

Previous studies have shown that either dominant-negative EHD1 (Lin et al., 2001) or EHD1-RNAi-knockdown (Naslavsky et al., 2004) cause accumulation of transferrin receptor at the ERC. To determine whether β1 integrins are similarly affected, we simultaneously compared the localization of internalized transferrin and β1 integrins in mock- and EHD1-RNAi-treated HeLa cells (Fig. 2). As expected, in mock-treated cells following the chase, transferrin was dispersed in small vesicles with only moderate levels of accumulation remaining at the ERC (Fig. 2A). The distribution of β1 integrins in these cells was also primarily a dispersed pattern with small punctate structures, showing partial overlap with transferrin-containing vesicles (Fig. 2A-C; arrows). However, EHD1-RNAi altered the subcellular distribution of internalized transferrin, inducing its accumulation at the ERC (Fig. 2D). In these same EHD1-RNAi-treated cells after chase, β1 integrins were also partially localized to a compact perinuclear region that showed partial overlap with the transferrin-containing structures at the ERC (Fig. 2D-F; arrows). Three-dimensional z-sections were obtained every 0.4 μm (sample sections shown in Fig. 2C,F), and the crosshairs and arrows illustrate that 10-20% of the transferrin and β1 integrins are colocalized on common structures within the ERC. Collectively, these data suggest that in HeLa cells EHD1 regulates the trafficking of internalized β1 integrins, possibly by controlling its exit from the ERC.

Integrins have been extensively studied in fibroblasts, and we next aimed to determine whether the loss of EHD1 in fibroblasts affected β1 integrin-mediated cellular functions. Accordingly, we first assessed the role of EHD1 in regulating β1 integrin trafficking in mouse embryonic fibroblast (MEF) cells derived from EHD1-knockout mice (Ehd1^–/– MEF) (Rapaport et al., 2006). Both the normal Ehd1^+/+ MEF and Ehd1^–/– MEF cells were subjected to fluorescence pulse-chase studies using 9EG7 anti-β1 integrin antibodies. In these experiments, the level of fluorescence visualized within the cells represents β1 integrins that have not recycled to the plasma membrane. As demonstrated in Fig. 3, under these conditions β1 integrins are internalized slowly and after a 1-hour pulse, much of the signal is still detected on the plasma membrane in both Ehd1^+/+ MEF and Ehd1^–/– MEF cells (Fig. 3A,B). Following a chase of 1 hour, the level of β1 integrins in Ehd1^+/+ MEF and Ehd1^–/– MEF cells is similar (Fig. 3C,D), but in many of the Ehd1^–/– MEF cells some accumulation is already noticeable in the ERC region (Fig. 3D). After 2 hours of chase, when compared with the Ehd1^+/+ MEF cells, there is a higher level of accumulation observed in the Ehd1^–/– MEF cells, particularly at the ERC (Fig. 3E,F; arrows). Upon 4 hours of chase, the Ehd1^+/+ MEF cells are mostly devoid of fluorescence, indicating that β1 integrin recycling is essentially complete (Fig. 3G). However, at this time point Ehd1^–/– MEF cells still retain detectable levels of β1 integrins within the cells (Fig. 3H). Indeed, quantification of β1 integrins retained in Ehd1^+/+ MEF and Ehd1^–/– MEF cells indicates an almost twofold difference in recycled β1 integrins at this time point, resulting from loss of EHD1 expression (Fig. 3I).

Since the four mammalian C-terminal EHD proteins are highly homologous and may interact with one another in protein complexes, we aimed to determine whether the other EHDs were also affected by loss of EHD1 expression, or whether the functional defects could be primarily attributed to loss of EHD1. As demonstrated in Fig. 3J, despite the loss of EHD1 expression in the Ehd1^–/– MEF cells, the levels of EHD2 and EHD4 were similar in both Ehd1^–/– MEF and normal MEF cells. Although it was difficult to assess changes in EHD3 (owing to a combination of low expression levels of EHD3 in MEF cells and/or the inability of the anti-human EHD3 antibody to efficiently react with mouse EHD3), loss of EHD1 appears to be the primary cause for the impaired β1 integrin recycling in the fibroblasts.

We next assayed human fibroblast cells by treating them with mock- or EHD1-RNAi (Fig. 3K-M). As demonstrated by immunoblotting, the efficacy of EHD1-knockdown was >90% (Fig. 3K). Although it is possible that the upper band of the doublet results from detection of EHD2 and/or EHD4 (which have a slightly higher molecular mass), our previous studies indicate that the EHD1 anti-peptide antibodies are specific for EHD1 (Naslavsky et al., 2006). Moreover, preliminary experiments suggest that this band is a phosphorylated form of EHD1 (M. Sharma and S.C., unpublished data). Having achieved significant reduction of endogenous EHD1 expression, we then subjected mock- and EHD1-RNAi-treated human fibroblasts to pulse-chase analysis with 12G10 anti-β1 integrin antibodies (Fig. 3L,M). In mock-treated cells most of the β1 integrin appeared to have recycled back to the plasma membrane within 2 hours of chase, with some staining remaining in intracellular vesicles (Fig. 3L). By contrast, at this time point, EHD1-RNAi-treated cells displayed a considerably higher level of integrin-containing punctate endosomal structures (Fig. 3M). Collectively, these data suggest that loss of EHD1 results in delayed recycling of β1 integrins to the plasma membrane.

β1 integrins play a crucial role in various cellular functions, including the regulation of focal adhesions, cell adherence to the extracellular matrix, cell spreading and cell motility (Hynes, 2002). To further appreciate the role of EHD1 and its impact upon β1 integrin trafficking and effects mediated downstream, we now sought to determine whether the cell-surface level of β1 integrin expression was altered in Ehd1^–/– MEF cells compared with Ehd1^+/+ MEF cells. Accordingly, Ehd1^+/+ MEF and Ehd1^–/– MEF cells growing either on plastic or FN were assayed for β1 integrins at the plasma membrane by flow cytometry experiments (Fig. 4A), using 9EG7 anti-mouse β1 integrin antibody that preferentially recognizes a ligand-bound conformational form of the integrin (Bazzoni et al., 1995). Since MEF cells are adherent, they were detached from non-coated or FN-coated plates by either brief trypsinization or cell scraping (Fig. 4A), or by brief incubation with EDTA (M.J. and S.C., unpublished data). In all cases, the level of activated β1 integrins at the plasma membrane was higher in the Ehd1^–/– MEF cells, generally displaying an increase of 50-100%. However, when β1 integrin surface levels were assayed using a conformation-independent MB1.2 antibody, we observed that the overall β1 integrin levels were decreased on Ehd1^–/– MEF cells (Fig. 4B). However, cell-surface levels of the lysosomal membrane protein LAMP1, which also localizes to the plasma membrane, were only slightly higher in Ehd1^–/– MEF cells (Fig. 4C). Epidermal growth factor receptor (EGFR) also showed a modest increase in MEF Ehd1^–/– cells at the plasma membrane, indicating that it too may be regulated through the EHD1 endocytic pathway (Fig. 4C). Overall, these results indicate that the loss of EHD1 leads to decreased β1 integrin levels on the plasma membrane, probably because of sequestering in an intracellular compartment. However, the increased localization of β1 integrins in an active conformation on the plasma membrane in Ehd1^–/– MEF cells raised the possibility of differences in focal adhesion dynamics in the absence of EHD1.

β1 integrins play an important role in the regulation of focal adhesion complexes. To study the potential impact of EHD1 upon focal adhesions, we analyzed and compared the localization of several focal adhesion constituent proteins in Ehd1^+/+ MEF and Ehd1^–/– MEF cells (Fig. 5). In Ehd1^+/+ MEF cells at steady state, focal adhesion kinase (FAK) was observed both in the cytosol as well as at the plasma membrane in focal adhesion plaques (Fig. 5A). However, Ehd1^–/– MEF cells displayed less FAK in the cytosol, with the majority of the protein localized to distinct focal adhesions at the plasma membrane (Fig. 5B). Another well-characterized focal adhesion protein, paxillin, also showed significantly more prominent and longer focal adhesion structures (as well as less cytosolic localization) in the Ehd1^–/– MEF cells compared with their normal MEF counterparts (Fig. 5C,D). A similarly altered subcellular localization was also observed using antibodies directed against the phosphorylated form of paxillin, also a key resident of focal adhesions (Brown et al., 2002) (Fig. 5E,F). Altered subcellular localization appeared to be the primary explanation for this phenomena, since neither enhanced expression nor increased phosphorylation of focal adhesion constituents was observed in Ehd1^–/– MEF cells (Fig. 5G). Other focal adhesion resident proteins, including vinculin and the tyrosine 397 phosphorylated form of FAK (Mitra et al., 2005), were also associated with more prominent focal adhesions in cells lacking EHD1 (M.J. and S.C., unpublished data).

An explanation for the prominent focal adhesions in Ehd1^–/– MEF cells might be a change in the kinetics of focal adhesion assembly and/or disassembly. To determine whether focal adhesion disassembly is impaired in Ehd1^–/– MEF cells, we synchronized the formation of focal adhesions and monitored their disassembly by an immunofluorescence-based assay (Ezratty et al., 2005; Franco et al., 2004). Ehd1^+/+ MEF and Ehd1^–/– MEF cells grown on FN-coated coverslips were incubated with nocodozole in the absence of serum, which induces the formation of focal adhesions through Rho-GTPase activity (Bershadsky et al., 1996; Liu et al., 1998) in addition to preventing microtubule polymerization. Serum-starved MEF cells displayed an elaborate microtubule network (Fig. 6A,B). Upon nocodozole treatment, the microtubules were disrupted and the tubulin distribution pattern was largely cytosolic for both Ehd1^+/+ MEF and Ehd1^–/– MEF cells (Fig. 6C,D). As the effect of nocodozole is reversible, we washed-out the drug and allowed microtubule polymerization to recover for 15-30 minutes (Fig. 6E-H). Having shown the efficacy of nocodozole and its reversible nature, we now turned to study the formation and disassembly of focal adhesions over time in this system, using paxillin as a representative component of focal adhesions. Serum-starved Ehd1^+/+ and Ehd1^–/– cells displayed weak focal adhesion staining and a mean focal adhesion length of ∼1 μm each (Fig. 6I,J,Q). The intensity of focal adhesion staining was greatly enhanced for both cell lines upon addition of nocodozole, with the mean focal adhesion length having doubled to approximately 2 μm (Fig. 6K,L,Q). Upon 15-30 minutes of nocodozole washout, Ehd1^+/+ MEF cells displayed a sharp reduction in both focal adhesion intensity and mean length because of their disassembly (Fig. 6M,O), approximating the focal adhesion intensity observed in starved, pre-nocodozole-treated cells (compare with Fig. 6I) and again displaying a mean focal adhesion length of ∼1 μm (Fig. 6Q). By contrast, the reduction in focal adhesion intensity and mean length following washout was markedly delayed in Ehd1^–/– MEF cells, with little change even after 30 minutes of washout (Fig. 6N,P,Q). These data suggest that focal adhesion disassembly is impaired in cells lacking EHD1 expression.

Integrin trafficking is crucial for the execution of integrin-mediated processes, such as adherence, cell spreading and motility (Caswell and Norman, 2006; Jones et al., 2006; Pellinen and Ivaska, 2006). Given that loss of EHD1 slows the dynamics of focal adhesion turnover in Ehd1^–/– MEF cells, we sought to determine whether integrin-mediated motility is influenced by the loss of EHD1. We compared the migration of Ehd1^+/+ and Ehd1^–/– MEF cells on FN- and VN-coated coverslips over 12 hours on a `scratch-assay' (Fig. 7). On FN, which engages β1 integrins, the Ehd1^+/+ MEF cells migrated into the scratched area, closing most of the wound and colonizing 65% of the wound area (Fig. 7C). At the same time, the Ehd1^–/– MEF cells colonized only 15% of the scratched region (Fig. 7D). When plated on VN, a ligand for αvβ3 integrins, both Ehd1^+/+ (Fig. 7E) and Ehd1^–/– MEF cells (Fig. 7F) showed similar degrees of colononization (26% and 22%, respectively). This level of migration on VN was significantly less than the 65% colonization achieved by Ehd1^+/+ MEF cells on FN, highlighting the role of EHD1 in regulating β1 integrin trafficking and downstream function. A previous study (Rapaport et al., 2006) and our current experiments (our unpublished observations) show the proliferation rates of Ehd1^+/+ MEF and Ehd1^–/– MEF to be very similar, rendering the enhanced wound colonization by Ehd1^+/+ MEF cells unlikely to be because of differences in cell growth within 12 hours. RNAi-knockdown of EHD1 in Ehd1^+/+ MEF cells also delayed migration, although the difference was not as marked (M.J. and S.C., unpublished data). To determine whether EHD1-knockdown alters the localization of β1 integrins on the cell surface of migrating cells, we compared the distribution of β1 integrins at the leading edge of Ehd1^+/+ and Ehd1^–/– MEF cells plated on FN (Fig. 7G,H). As shown, the Ehd1^+/+ cells displayed a large number of β1 integrin-containing filopodial spikes that projected from the plasma membrane towards the wound area (Fig. 7G and inset). By contrast, Ehd1^–/– MEF cells exhibited few filopodia extending towards the wound, but displayed levels of integrins at the plasma membrane as well as being particularly pronounced in intracellular structures (Fig. 7H and inset). Overall, these findings define a role for EHD1 in regulating integrin-mediated cell motility, and suggest that the function of β1 integrins is more tightly regulated by EHD1-knockdown than that of β3 integrins.

To further examine the functional significance of EHD1-mediated control of β1 integrin trafficking, we compared the ability of Ehd1^+/+ MEF and Ehd1^–/– MEF cells to spread on FN (Fig. 8). We calculated that 80 randomly chosen Ehd1^+/+ MEF cells yielded a mean surface area of ∼2300 μm² after 3 hours of spreading (Fig. 8A,E). By contrast, the same number of Ehd1^–/– MEF cells spread less at this time point, and their mean surface area was calculated at only approximately 1500 μm² (Fig. 8B,E). Differences of up to threefold in spreading were observed in individual experiments, depending on the time the cells were allowed to spread prior to fixation. Upon introduction of a GFP-EHD1 construct into Ehd1^–/– MEF cells, the rate of cell spreading was partially rescued (Fig. 8C,E; green denotes transfected cells) when compared with neighboring non-transfected cells (red-only cells). Lack of complete rescue probably results from heterogeneous expression levels of the transfected GFP-EHD1 in the Ehd1^–/– MEF cells. As an additional control, we used a GFP-EHD1 mutant with a point mutation in a conserved glycine residue (EHD1 G65R) crucial for nucleotide binding (Lee et al., 2005; Naslavsky et al., 2006) and localization of EHD1 to membranes, rendering the mutant EHD1 dysfunctional (Caplan et al., 2002; Grant et al., 2001; Lin et al., 2001). As shown, GFP-EHD1 G65R was incapable of rescuing cell spreading in Ehd1^–/– MEF cells (Fig. 8D,E), and even showed a modest additional decrease in cell spreading. Collectively, these data support a role for EHD1 in regulating β1 integrin trafficking and downstream functions, such as cell spreading and migration.

Despite the crucial functions that integrins play in modulating the interaction of the cell with the extracellular matrix, until recently relatively little emphasis has been placed on understanding the subcellular itinerary of these proteins and determining the mode by which this itinerary is regulated. Given the importance of β1 integrin trafficking for a wide variety of cell functions, especially motility and migration, tumor cell invasiveness (reviewed by Caswell and Norman, 2006) and cell adhesion and spreading (Dunphy et al., 2006; Roberts et al., 2001), renewed focus is being placed on the identification of proteins involved in the regulation of integrin trafficking. We have demonstrated that the C-terminal EHD protein EHD1 plays an important role in modulating β1 integrin function by regulating recycling of these receptors.

Initially, using HeLa cells as a well-established model to study alterations in the subcellular itinerary of β1 integrins (Powelka et al., 2004), we demonstrated that RNAi-knockdown of EHD1 led to an accumulation of β1 integrins in a transferrin-containing perinuclear region corresponding to the ERC. Moreover, we detected membrane vesicles containing both internalized β1 integrins and EHD1. Accordingly, EHD1, whose primary role appears to be in regulating the return of endocytosed cargo to the plasma membrane from a Rab11-positive recycling compartment (Caplan et al., 2002; Grant et al., 2001; Lin et al., 2001; Naslavsky et al., 2004; Naslavsky et al., 2006; Rapaport et al., 2006), may play a similar and/or overlapping role to that of Rab11 in regulating the long loop of β1 integrin recycling (Roberts et al., 2001). Accumulation of intracellular β1 integrins was also documented upon RNAi-knockdown of the EHD1 binding partner, Rabenosyn-5 (M.J. and S.C., unpublished data), similar to the accumulation observed for internalized transferrin (Naslavsky et al., 2004). Since it has recently been demonstrated that Rab21 and Rab5 associate with integrins (Pellinen et al., 2006), and Rabenosyn-5 is a divalent effector of both Rab4 and Rab5 (Nielsen et al., 2000) (and possibly also Rab21), one might speculate that integrin endocytosis may be coordinated jointly by GTP-binding proteins such as Rab21/Rab5 and the C-terminal EHD1 protein through interactions with common effectors such as Rabenosyn-5. Although the precise mechanism of Rabenosyn-5 function is not clear, it has been implicated in vesicle fusion (Nielsen et al., 2000). Interestingly, integrins such as α5β1 also accumulate in a Rab11-containing compartment when SNARE activity is inhibited (Skalski and Coppolino, 2005), reinforcing the notion that SNARE-mediated fusion is important in the regulation of integrin recycling.

Although HeLa cells served as an established model for the initial assessment of EHD1 in β1 integrin trafficking (Becker and Hannun, 2003; Powelka et al., 2004), integrin function has been studied more extensively in fibroblasts. Using MEF cells derived from mice lacking EHD1 expression (Rapaport et al., 2006), we examined the role of EHD1 in regulating the trafficking of β1 integrins. Loss of EHD1 led to an accumulation of intracellular β1 integrins observed after 2 hours of chase, when compared with β1 integrins in normal Ehd1^+/+ MEF counterpart cells. Moreover, after 4 hours of chase, an approximately twofold increase was observed in the level of non-recycled β1 integrins. The role of EHD1 in regulating β1 integrin recycling was further supported by flow cytometry experiments with the MB1.2 antibody (which recognizes conformation-independent β1 integrins), demonstrating that overall levels of β1 integrins are reduced on the cell surface as a result of impaired recycling (Fig. 4B). To further validate the role of EHD1 in β1 integrin recycling, an RNAi approach was applied to a human fibroblast cell line and yielded similar results. Morphologically, the accumulation of β1 integrins at the ERC in the human fibroblasts had a more dispersed pattern than in HeLa and MEF cells. Indeed, the ERC in the human fibroblasts is not as well defined and not relegated exclusively to the perinuclear region as observed in HeLa, CHO and some other cell types. These data provide evidence promoting a role for EHD1 in the regulation of β1 integrin trafficking.

To transduce signals into the cell, integrins localized to the plasma membrane first need to recruit a series of intracellular proteins and assemble them into small plaques known as focal complexes or focal adhesions (Wozniak et al., 2004). Recent studies have shown that focal adhesion formation is a complicated process, entailing the sequential recruitment of multiple proteins with different kinetics (Webb et al., 2003). Focal adhesion disassembly involves microtubules, dynamin and FAK (Ezratty et al., 2005), and the protease calpain (Franco et al., 2004). Our data demonstrate that Ehd1^–/– MEF cells exhibit more prominent focal adhesions [determined by staining for focal adhesion constituent proteins such as paxillin (Brown et al., 2002; Schaller, 2001; Turner, 2000) and FAK (Mitra et al., 2005)], resulting from slower disassembly kinetics. Interestingly, similar to the loss of EHD1 expression, interference with the function of the small GTPase Rab21 (by use of a dominant-negative GDP-locked mutant) also induced prominent focal adhesions (Pellinen et al., 2006). Although it is not understood how the absence of EHD1 stabilizes focal adhesions, our findings that there is no enhancement in the levels of phosphorylated paxillin and phosphorylated FAK in Ehd1^–/– MEF cells, along with preliminary data suggesting that calpain activity is not affected in Ehd1^–/– MEF cells (M.J. and S.C., unpublished data), support the notion that impaired recycling of β1 integrins or one of the many factors involved in focal adhesion disassembly may be responsible for these phenomena. Moreover, the increased levels of active β1 integrins detected at the plasma membrane of the Ehd1^–/– MEF cells (Fig. 4A) may also be related to the delay in focal adhesion disassembly.

The recycling of integrins as well as assembly and disassembly of focal adhesions are crucial for polarized cell motility (Webb et al., 2004) (reviewed by Wozniak et al., 2004). Cell motility depends on the internalization of integrins at the rear of a migrating cell and their subsequent recycling to the plasma membrane at the leading edge of the cell (Bretscher, 1992; Lawson and Maxfield, 1995; Strachan and Condic, 2004). It has been further suggested that this activity involves elaborate interactions between Arf6 and ArfGAP complexes that are proposed to recruit paxillin and other focal adhesion constituents to newly assembling focal adhesions containing Rac-induced microfilaments at the leading edge of the cell (reviewed by de Curtis, 2001). Indeed, a recent report demonstrates that RNAi-knockdown of ACAP1, a GTPase-activating protein (GAP) for Arf6 (Jackson et al., 2000), inhibits β1 integrin recycling and impairs cell migration on a FN-coated matrix (Li et al., 2005). In accordance with these findings, another study shows that loss of expression of the Arf6 GTP exchange factor (GEF), GEP100/BRAG2, causes enhanced cell attachment and spreading on FN-coated substrates (Dunphy et al., 2006). Further evidence of Arf6 involvement in the regulation of integrin trafficking is supported by the recent finding that recycled β1 integrins colocalize with Arf6 in peripheral membrane ruffles (Powelka et al., 2004). We have previously shown a substantial colocalization between Arf6 and EHD1, and that the subcellular distribution of EHD1 is dramatically altered in response to the nucleotide status of Arf6 and to the expression of both ACAP1 and EFA6 (Caplan et al., 2002), Arf6 GAPs and GEFs, respectively. It is noteworthy that both the stabilization of focal adhesions and the significant decrease observed in the rate of fibroblast migration in fibroblasts lacking EHD1 in this study are similar to the effects caused by transfection of a dominant-negative Rab21 (Pellinen et al., 2006). In addition, the combination of loss of polarized filopodia and delayed migration that we observed in Ehd1^–/– MEF cells when grown on FN is reminiscent of the effects documented for FAK^–/– MEFs (Ilic et al., 1995).

Although it is presently unclear why delayed β1 integrin recycling stabilizes focal adhesions, the most likely possibility is that the trafficking of some of the proteins playing key roles in focal adhesion disassembly is regulated through an EHD1-mediated recycling pathway. Potential examples include FAK, Src, p130CAS, extracellular signal-related kinase (ERK) and myosin light chain kinase (Webb et al., 2004) or Endo180 (Sturge et al., 2006). Fibroblasts lacking EHD1 exhibit a considerable delay in their spreading on a FN-coated surface, and this phenotype can be rescued by the introduction of wild-type EHD1, implying that this loss of protein function is also a primary cause of defective cell spreading.

Our data point to a role for EHD1 in the control of β1 integrin trafficking, adding this endocytic protein to the growing list of proteins that regulate the subcellular itinerary of β1 integrins. The precise mechanisms of the complex physical and functional interactions between EHD1, Rab4, Rab11, Arf6 and their effectors remain to be elucidated, and a better understanding of their coordinated control of endocytic transport will shed new light on the trafficking and function of integrins.

All cDNA constructs have been described previously (Caplan et al., 2002; Naslavsky et al., 2004). The following antibodies were used: 12G10 mouse monoclonal antibodies to human β1 integrins, with higher affinity to the activated conformation (Mould et al., 1995; Powelka et al., 2004) (cat. # MCA2028, Serotec, Oxford, UK), LAMP1 (Santa Cruz Biotechnology, Santa Cruz, CA), β-actin (Novus Biologicals, Littleton, CO), α-tubulin (Invitrogen, Carlsbad, CA), FAK (Upstate, Lake Placid, NY), paxillin and EGFR (BD Transduction Laboratories, Franklin Lakes, NJ), rabbit phospho-specific antibodies to paxillin phosphorylated at Y118 and FAK phosphoryated at Y397 (Biosource, Camarillo, CA), rat 9EG7 monoclonal antibodies that preferentially recognize ligand-bound mouse β1 integrins (Bazzoni et al., 1995) (cat. # 550531, BD Pharmingen, San Diego, CA), and conformation-independent MB1.2 rat monoclonal antibodies against mouse β1 integrins (Sakai et al., 1998; Von Ballestrem et al., 1996) (cat. # MAB1997, Chemicon, Temecula, CA). Affinity-purified rabbit polyclonal antibodies against human EHD1 (DLPPHLVPPSKRRHE), EHD2 (VERGPDEAMEDGEEGSD) and EHD4 (SHRKSLPKAD) (AnaSpec, San Jose, CA) were prepared. Cy-3-conjugated anti-mouse and anti-rabbit immunoglobulin G (IgG), Alexa Fluor 488-conjugated antibody to mouse and rabbit IgG, Alexa Fluor 633 goat anti-mouse, Alexa Fluor 633 goat anti-rat, Alexa Fluor 568 goat anti-rat, transferrin–Alexa-Fluor-568 and Cy-3-labeled wheat germ agglutinin (WGA) were from Invitrogen. Goat anti-mouse horseradish peroxidase (HRP) and donkey anti-rabbit HRP were obtained from Jackson ImmunoResearch (West Grove, PA).

Ehd1^–/– MEF cells were previously described (Rapaport et al., 2006). Human foreskin-derived fibroblasts were obtained from ATCC (BJ; CRL-2522, Manassas, VA). MEF and HeLa cells were maintained in Dulbecco's modified Eagle's medium (DMEM) (high glucose), whereas human foreskin fibroblasts were maintained in Eagle's minimal essential medium (EMEM). All of the media was supplemented with 10% fetal bovine serum (FBS), 100 μg/ml streptomycin, 100 U/ml penicillin and 2 mM L-glutamine.

Oligonucleotide duplexes targeting human EHD1 (bp 942–963: AAGAAA-GAGATGCCCAATGTC, Dharmacon, Lafeyette, CO) or mouse EHD1 (bp 921–942: AAGACATCCAGTCTCTGCCGA, Dharmacon) were transfected into HeLa cells or Ehd1^+/+ MEF cells for 48 hours using Oligofectamine (Invitrogen, Carlsbad, CA) (Elbashir et al., 2001). SMARTpool oligonucleotide duplexes (Dharmacon), targeting human-EHD1, were transfected into human foreskin fibroblasts using Dharmafect (Dharmacon) for 48 hours. Efficacy of EHD1-RNAi was confirmed by immunoblotting lysates with antibodies for EHD1.

HeLa and human fibroblast cells were grown on coverslips, and transfected with appropriate EHD1-RNAi oligonucleotides for 48 hours. Based on a modified protocol of Hsu and co-workers (Powelka et al., 2004), cells were starved for 2 hours at 37°C in DMEM lacking serum [but containing 0.5% bovine serum albumin (BSA)], followed by a 2-hour pulse with 5 μg/ml antibody to human β1 integrin at 37°C. Surface antibodies were removed by an acid rinse (0.5% acetic acid, 0.5 M NaCl, pH 3.0) for 45 seconds. The chase was performed in complete media (containing 10% FBS) at 37°C for 2 hours and cells were fixed with 4% (v/v) paraformaldehyde (PFA) in phosphate-buffered saline (PBS). Ehd1^+/+ MEF cells and Ehd1^–/– MEF cells were processed in similar fashion, using a 1-hour pulse and 1-4 hours of chase.

For combined β1 integrin/transferrin recycling assays, at the end of the 2-hour pulse with β1 integrin antibodies, the cells were then pulsed for 10 minutes with transferrin–Alexa-Fluor-568, and chased for 15 minutes longer in complete media at 37°C prior to fixation. Fixed cells were then incubated with the appropriate fluorochrome-conjugated secondary antibody mixture containing 0.1% (w/v) saponin and 0.1% (w/v) BSA for 30 minutes. All images were acquired on an LSM 5 Pascal confocal microscope (Carl Zeiss, Thornwood, NY) by using a 63×1.4 numerical aperture objective with appropriate filters. For combined internalization studies with both internalized transferrin and β1 integrins, 4 μm z-sections were obtained and colocalizations were validated in three-dimensions.

To quantify β1 integrin recycling by flow cytometry, HeLa cells at sub-confluent density were treated with EHD1-RNAi or mock-treated. After 48 hours the cells were serum-starved for 2 hours, pulsed with anti-β1 integrin antibody for 2 hours, acid-stripped and chased in complete media for 2 hours. Following trypsinization and pelleting, cells were incubated under non-permeabilizing conditions with Alexa Fluor 633-conjugated goat anti-mouse secondary antibody for 30 minutes, and fixed in 4% PFA/PBS. At least 10,000 cells were analyzed for reappearance of β1 integrins on the plasma membrane by flow cytometry analysis (BD Biosciences, San Jose, CA).

MEF cells were serum-starved in DMEM with 0.5% BSA on glass coverslips for 72 hours prior to addition of 10 μM nocodazole for 15 minutes at 37°C, and either fixed directly or first incubated in starvation media (Ezratty et al., 2005). Cells were then immunostained with either mouse monoclonal antibodies against paxillin or tubulin and appropriate secondary antibodies. Quantification of mean focal adhesion length was done using LSM 5 Pascal software for an average of 227 adhesions per treatment. One-tailed Student's t-tests were applied to test statistical significance of the data.

MEF cells were grown at equivalent confluence for 18 hours on coverslips precoated with 10 μg/ml FN or 16.7 μg/ml VN. A single scratch of ∼700 μm was introduced with a pipette tip, and cells were allowed to migrate into the wound for 12 hours at 37°C. Cells were visualized by immunostaining with fluorochrome-labeled WGA and confocal microscopy using a 10×0.3 numerical aperture objective.

MEF cells were trypsinized and replated on coverslips precoated with 10 μg/ml FN, and allowed to spread for 3 hours at 37°C in the presence of complete media. Cells were fixed and labeled with Cy-3-conjugated WGA for 30 minutes to better visualize the cell outlines. Cell surface boundaries were outlined for 80 individual cells chosen randomly and LSM 5 Pascal software was used to calculate the mean surface area and standard deviation of each population. One-tailed Student's t-tests were applied to test the statistical significance of the data.

We are grateful to M. Wheelock and K. Johnson and their group for critically reading the manuscript and extremely helpful discussions. We thank Linda Wilkie and the UNMC flow cytometry core facility. M.J. was supported by an INBRE/BRIN fellowship from National Institutes of Health Grant P20 RR16469. This work was supported by the National Institutes of Health Grants P20 RR018759 (S.C.) from the National Center for Research Resources and 1R01 GM074876-01 (S.C.) from the National Institute for General Medical Sciences.