Rho/ROCK and myosin II control the polarized distribution of endocytic clathrin structures at the uropod of moving T lymphocytes

Cell polarity is the inherent ability of most cells to create morphological asymmetry and to separate regions involved in specialized functions (Drubin and Nelson, 1996; Wedlich-Soldner and Li, 2003). A well-known example occurs during directional migration; distinct molecular processes occur at the front and back of a moving cell (Ridley et al., 2003). This is particularly evident in fast-moving cells, such as leukocytes or the soil amoeba Dictyostelium discoideum, which exhibit a strong cell polarity displaying a leading pseudopod, the mid cell body containing the nucleus and the trailing uropod (Sanchez-Madrid and del Pozo, 1999; Van Haastert and Devreotes, 2004).

A simple mechanism for inducing cell asymmetry is membrane recycling: locally exocytosed proteins will remain polarized if they are endocytosed and recycled before they can diffuse to equilibrium (Bretscher, 1996). Genes that prescribe landmarks in the plasma membrane for polarized secretion, and mechanisms that recognize and reinforce the landmark are well characterized in yeast (Nelson, 2003). Whether a second landmark might signal to polarize endocytosis has remained unknown. In epithelial cells, endocytosis occurs at both the apical and basolateral surfaces of the cell and, subsequently, polarized sorting in endosomes helps to maintain distinct domains (Mellman, 1996; Traub and Apodaca, 2003). In migrating cells, the exocytic and endocytic pathways are less well characterized; early works described polarized exocytosis of vesicles toward the leading edge of spreading cells (Bretscher, 1983; Hopkins et al., 1994). Consistent with these studies, direct visualization of a circulating receptor exocytosis showed polarization towards the leading edge (Schmoranzer et al., 2003). It has been proposed that cell adhesion receptors, like integrins, are preferentially delivered toward the leading lamella whereas, at the cell rear, integrins are internalized (Webb et al., 2002).

The major route of plasma membrane protein endocytosis is the clathrin-mediated pathway, characterized by the selective internalization of receptors via clathrin-coated vesicles (CCVs) (Conner and Schmid, 2003; Sorkin, 2004). To form CCVs at the plasma membrane, structural proteins, including clathrin and adaptor proteins (AP-2), are recruited to clathrin-coated pits (CCPs), which detach from plasma membrane as CCVs in a dynamin-dependent process (Keen, 1990; Perrais and Merrifield, 2005). With regard to cell migration, there is currently no evidence that endocytic rates are augmented at the cell rear (Jones et al., 2006). Indeed, the localization of clathrin towards the leading and not the trailing edge of migrating adherent cells has been demonstrated (Rappoport and Simon, 2003). By contrast, in fast moving Dictyostelium, clathrin-GFP is transiently concentrated on the membrane at the back of the cell tail (Damer and O'Halloran, 2000); similarly, in polarized leukocytes, coated pits and vesicles are concentrated at the uropod (Davis et al., 1982).

We have studied the spatio-temporal dynamics of clathrin-mediated endocytosis (CME) using live-cell fluorescence microscopy to visualize clathrin and AP-2 structures during the internalization of a cargo molecule (transferrin; Tfn) in moving T lymphocytes. Our results show a polarized distribution of endocytic structures towards the uropod during lymphocyte migration, which is under the control of Rho/Rho-kinase (ROCK) and myosin II function. Importantly, the specific disruption of CME reduced the migratory ability of T lymphocytes.

Migrating T cells are spontaneously polarized cells that extend a unique structure at their back called the uropod (del Pozo et al., 1996). To study the role of CME, we transfected T cells with GFP-tagged clathrin-light chain, which fully co-localized with the endogenous clathrin-heavy chain (supplementary material Fig. S1A). We found that most polarized T cells showed clathrin enrichment at their backs (Fig. 1A). The posterior localization of the Golgi network in polarized T cells probably accounts for much of the intracellular GFP-clathrin fluorescence detected within the uropod (supplementary material Fig. S1B), but these structures were excluded from our analysis since transferrin does not transit through them (see below). We were interested in a distinct subset of clathrin structures (CSs) detected in an area adjacent to the plasma membrane that could be involved in CME. Interestingly, CSs appeared frequently clustered in certain regions of the uropod membrane of living T cells expressing GFP-clathrin (Fig. 1A). Compared with other cell types that displayed abundant GFP-clathrin spots at the adherent surface, polarized T lymphocytes attached to fibronectin or to ICAM-1 showed almost no significant clathrin spots at their lower surface (Fig. 1A and supplementary material Fig. S1C). Because the uropod is usually elevated above the adhesion plane, CSs were imaged in confocal cross-sections of the plasma membrane. A heterogeneous population of membrane-associated clathrin arrangements were detected; some diffraction-limited spots (∼200 nm) (and supplementary material Fig. S1D), likely corresponding to single CCPs, and a major population of larger CSs (300-800 nm); which displayed distinct morphologies with multiple curved domains or relatively flat areas (Fig. 1A and supplementary material Fig. S1E). We analyzed in detail several aspects of CSs; first, we confirmed their sub-membrane position co-staining the plasma membrane with anti-CD45 (supplementary material Fig. S1E). Second, to assess the endocytic nature of CSs, T cells were briefly incubated with the clathrin-dependent cargo transferrin labelled with Texas Red (Tfn-Tx). Smaller and uniform diffraction-limited Tfn spots (<200 nm) were frequently detected close to CSs, which may correspond to CCPs or budding CCVs (Fig. 1A, arrows; note the small spot of Tfn/clathrin budding from the side of a large CS in the magnified image). Third, to discriminate between plasma membrane CSs and a clathrin-coated endosomal population that might be close to the cell surface, we also studied the distribution of the endocytic clathrin adaptor AP-2, using a specific anti-α-adaptin Ab. Similarly to CSs, the plasma membrane adaptor AP-2 was highly polarized toward the uropod, both in primary T cells and in GFP-clathrin expressing cells, where it partially co-localized with most membrane CSs (Fig. 1B,C). The polarized distribution of CSs was clearly observed in GFP-clathrin fluorescence intensity profiles along the cell perimeter (Fig. 1D). Polarization was also assessed by screening different membrane domains (Fig. 1E). We visualized >400 CSs labeled with GFP-clathrin or AP-2 from 49 confocal sections (n=26 cells), showing that 62±7% were located along the rear uropod membrane, 27±4% along the uropod neck and 11±2% at the leading edge. Histograms in Fig. 1F summarize the frequency of membrane CSs/μm. Collectively, these data suggest that different-sized CSs, which may be involved in CME, are clustered at the uropod membrane of polarized T cells.

Recent works suggest that large CSs correspond to platforms for cargo selection, supporting several rounds of vesicle formation (Perrais and Merrifield, 2005). Thus, we focused on the temporal dynamics of large CSs with respect to cargo internalization (Fig. 2A,B). Large CSs remained stable throughout image acquisition and it was possible to monitor more than one diffraction-limited Tfn spot transiently co-localizing at a given CS, which were visualized in cells expressing either GFP-clathrin or GFP-α-adaptin (Fig. 2A,B, arrowheads). During cargo internalization, Tfn was steadily accumulated at endosomes that occasionally were detected very close to plasma membrane CSs (Fig. 2A, arrows). Spatio-temporal projections of later times allowed the identification of numerous co-localizing events concentrated around larger Tfn-positive (Tfn⁺) endosomes located deeper in the cytosol (Fig. 2C). These large Tfn⁺ endosomes partially co-localized with a heterogeneous population of cytoplasmic clathrin in a transient and dynamic manner, which probably corresponds to cargo sorting processes (Fig. 2D). Thus, both CCP assembly at different-sized CSs, and the later traffic of CCVs from the membrane and through endosomes, appear to be polarized at the uropod of T cells.

To analyze Tfn internalization during net cell locomotion, we visualized free-moving cells plated on ICAM-1. Upon exposure to Tfn, it was initially bound all around the cell, including the leading lamellipodium (Fig. 3A, time 0 seconds). After following the random migration through a distance equivalent to several cell diameters, an important accumulation of Tfn in the posterior membrane and endosomes was observed (Fig. 3A and supplementary material Movie 2). This observation prompted us to follow the distribution of Tfn binding to its receptor (TfnR). TfnR is surface expressed all over the plasma membrane, including the lamellipodium (Fig. 3B), and intracellularly accumulated at recycling endosomes (not shown). Immediately after a Tfn-A488 pulse, there was a complete co-localization between the ligand and its receptor all over the cell surface, and almost no Tfn was detected intracellularly (Fig. 3B). After 4 minutes of chase, Tfn only co-localized with membrane TfnR at the uropod, whereas most Tfn was already internalized (Fig. 3C).

To demonstrate the in vivo displacement of Tfn-TfnR complexes on the membrane, we next performed bleaching of membrane-bound Tfn in primary T lymphocytes. Cells were chilled to block CME and surface stained with Tfn, which was washed off before warming up the cells. Soon after warming to 37°C, approximately 30% of the cells attached to fibronectin displayed recognizable pseudopods and uropods in opposite cell poles, allowing the bleaching of the plasma membrane corresponding to either the anterior or the posterior part of the cell. Bleaching the posterior pole, we found that approximately 43±18% (n=6) of original Tfn-fluorescence was recovered at the bleached membrane, which shows an active diffusion of Tfn at the membrane (Fig. 4A). This result is in the range of the expected recovery for a pool-limited molecule; actually, the distribution of Tfn on the membrane 300 seconds after bleaching was almost homogeneous, and so the intensity ratio between the bleached uropod and the front membranes was close to one (Fig. 4B). Interestingly, much lower recovery (5±2%, n=5) was detected after bleaching the front membrane (Fig. 4C). The existence of a rearward displacement of Tfn-TfnR complexes might explain these differences. Anti-CD45-FITC was included as a control membrane protein that is not internalized, and its fluorescence was not recovered after bleaching any pole (Fig. 4B). Moreover, Tfn uptake was almost unaffected (91±4% of control cells, n=5) when bleaching the front membrane at short time-lapses, which shows that bleaching did not affect CME (Fig. 4C). However, rapid Tfn internalization was not detected in rear-bleached cells compared with neighbor control cells in the same time-lapse (4±6%, n=6), which suggests a `bleached' membrane origin for Tfn (Fig. 4A,D and supplementary material Movie 3). Together these data favor a polarized model of Tfn internalization (cargo bound to its receptor flows through the membrane from the anterior to the posterior pole prior to being internalized), versus a general model of internalization all over the cell and subsequent cytoplasmic vesicle traffic towards the polarized endosomes (Figs 3, 4).

The directed flow of Tfn-TfnR complexes towards the uropod may be a sign of the process known as retrograde cortical flow that depends on myosin II contractility (Bray and White, 1988). To explore the role of myosin II on CME polarization, we used either the specific myosin II ATPase inhibitor blebbistatin or blocked myosin II assembly through inhibition of ROCK with Y-27632, or Rho by the expression of C3 transferase or dominant negative RhoA (RhoAN19). Upon Rho/ROCK/myosin inhibition, cells lose their rigidity, the uropod lays down on the coverslip and the whole cell flattens. In addition, cells plated on integrin ligands display an unusual polarized phenotype; the cell body develops a strong protrusive activity and forward movement, leaving a long unretracted tail behind (Fig. 5B, Fig. 6). The latter shape is similar to that previously reported in leukocytes treated with Y-27632 (Alblas et al., 2001; Rodriguez-Fernandez et al., 2001; Worthylake et al., 2001). We analyzed the effects of these treatments on short-term Tfn uptake (5-10 minutes). Blebbistatin treatment produced a kinetic delay and partial inhibition of endocytosis; cells internalized 50-60% of Tfn compared with controls (Fig. 5A). Similar results of partial inhibition were obtained with Y-27632 or C3-GFP, and with the dominant-negative RhoAN19-GFP (Fig. 5A). ROCK and myosin II blockade did not affect clathrin and AP-2 assembly, or their association, as shown by co-immunoprecipitation (supplementary material Fig. S2A). Disruption of actin with cytochalasin D or latrunculin B resulted in total inhibition of endocytosis in approximately 75% of cells (Fig. 5A). Following the removal of blebbistatin, cytochalasin D or latrunculin B, Tfn endocytosis was rapidly restored (Fig. 5A). Colchicine had no effect on Tfn endocytosis, nor on CS or AP2 rear distribution (Fig. 5A and supplementary material Fig. S2C,D). Interestingly, in Rho/ROCK/myosin-II-inhibited cells, some Tfn was still on the membrane (co-localizing with surface-TfnR) and internalized Tfn was already accumulated in dispersed endosomes throughout the cell body (Fig. 5B and supplementary material Fig. S3). We next analyzed the effects of the inhibitors on the subcellular distribution of CSs (Fig. 6). Notably, plasma membrane endocytic structures labeled with GFP-clathrin, endogenous clathrin or AP-2, were observed all over the cell surface, and easily detected and quantified at the basal membrane (Fig. 6 and supplementary material Fig. S2B,E; see high magnification panels showing CSs at the lower membrane surface, note the low frequency of CSs in control cells). Thus, Rho/ROCK signaling and myosin II contractility appear to control the posterior position of CME in polarized T lymphocytes.

To analyze the role of CME in cell migration, we used the dominant-negative DIII construct of Eps15 that prevents AP-2 membrane recruitment, and the negative control GFP-D3Δ2 lacking all the AP-2-binding sites (Benmerah et al., 1998). Three distinct T-cell lines (HSB-2, CEM and PEER) were transiently transfected with these constructs, and tested for their capacity to migrate in chemotaxis chamber assays towards the chemotactic stimuli CXCL12 (SDF1). As expected, cells expressing GFP-DIII failed to internalize Tfn, compared with neighboring untransfected cells or with cells transfected with control GFP-D3Δ2 or with GFP alone, which internalized Tfn similarly to untransfected cells (Fig. 7A, Fig. 8). Importantly, cells overexpressing the dominant-negative inhibitor of constitutive endocytosis GFP-DIII were less efficient in migrating to the lower compartment of the chemotaxis chamber compared with cells transfected with control GFP-D3Δ2 or with GFP alone (Fig. 7B). GFP-DIII cells, in which Tfn endocytosis was completely inhibited, displayed no changes in cell morphology (polarization, uropod length, roundness and area) and were able to cap ICAM-3 at the uropod (Fig. 8A). GFP-DIII cells bound Tfn all around the cell surface with no apparent redistribution towards the uropod during 15 minutes of video capture (Fig. 8B). The expression of this mutant appears to impede transferrin rearwards flow and/or its back retention, as shown by measurements of Tfn fluorescence intensity profiles along membranes (Fig. 8C).

To further analyze the role of clathrin in cell migration, we performed RNA interference to downregulate clathrin gene expression with three commercially available small inhibitory RNAs (siRNAs). Two clathrin-siRNA transfectants exhibited blockade of clathrin expression, which was accompanied by a failure in Tfn endocytosis (Fig. 7C,D). Clathrin-siRNA transfectants exhibited a notable impairment in both basal and CXCL12-induced migration (Fig. 7E,F).

In this work we thoroughly dissected the process of CME in the highly polarized and dynamic cell model of migrating T cells. We show the unique redistribution pattern of clathrin and other components of the CME, such as AP-2, towards the posterior uropod. Various-sized CSs are clustered at the posterior plasma membrane: some diffraction-limited spots, and a major population of larger and static CSs. Our dynamic studies during cargo internalization suggest that large CSs are endocytic active platforms that support multiple rounds of endocytic vesicle creation. We also show that this posterior distribution is governed by Rho/ROCK and myosin II. Finally, we have analyzed the functional consequences of CME disruption in T lymphocytes and we show that cell migration is severely attenuated.

Most dynamic CME studies recognize productive endocytic events as the membrane-associated GFP-clathrin spots that disappear from the field of observation, when viewing the basal adherent surface (the inward movement of CCP/CCV occurs along the optical z-axis) (Rappoport et al., 2004). This on-off behavior of diffraction-limited CSs fits well with the classical model of CME that begins with the nucleation and growth of one CCP at a site on the membrane and is followed by the invagination and detachment of one CCV into the cytosol (Ehrlich et al., 2004). However, CSs also include a population of larger and more complex structures that may support multiple rounds of vesicle formation before being disassembled and correspond to platforms for cargo selection (Merrifield et al., 2005; Perrais and Merrifield, 2005). Our results show that distinct-sized CSs are present at the plasma membrane of polarized T lymphocytes, including both small and large CSs. The dynamic identification of endocytic cargo inside large clathrin and AP-2 structures (transient co-localization with small Tfn spots) provides a formal proof of their endocytic ability. These results are in line with recent studies indicating the existence of endocytic `hot platforms' that may increase the speed and efficiency of CME in other cell types such as primary adipose cells (Bellve et al., 2006). Indeed, the mechanisms by which endocytosis initiates at particular locations on the plasma membrane are beginning to be elucidated in yeast; new structures, called `eisosomes', mark hotspots of endocytosis (Walther et al., 2006). Whether selective endocytic zones at the T-cell uropod could resemble eisosome-like structures in higher eukaryotes deserves further investigation.

Interestingly, we show that most endocytic CSs, including small and large ones, are clustered toward the uropod surface of polarized T lymphocytes. In a previous study with migrating cells, an enhancement of GFP-clathrin at the basal surface toward the leading edge was detected with total-internal-reflection fluorescence microscopy (TIR-FM), which looks only at those fluorescently labeled organelles within 100 nm of the lower plasma membrane (Rappoport and Simon, 2003). Nevertheless, our differences may be explained by the distinct kind of cell migration: slow and adherent mesenchymal type versus rapid amoeboid movement. It is possible that only cells displaying a rear uropod can concentrate CME toward this unique structure, whose function is not yet well defined. In this work we analyzed the presence of membrane-associated CSs in confocal sections of the whole cell, not only at the basal surface. We detected most membrane-associated CSs clustered in cross-sections of the uropod membrane, where the inward movement of CCP/CCV occurs on the x,y plane instead of `disappearing' along the optical z-axis. Therefore, we visualized simultaneously the membrane and the cytosol during Tfn internalization, and found that most membrane and some cytoplasmic CSs, which probably correspond to clathrin-coated-endosomes, co-localized with Tfn. The spatio-temporal sequence of clathrin/Tfn and AP-2/Tfn co-localization allowed their identification: large endocytic structures are clathrin or AP-2 static structures at the membrane that transiently interact with spots of Tfn, whereas clathrin-coated-endosomes are AP-2-negative and steadily accumulate Tfn. Other larger endosomes that accumulate Tfn at later times display transient interactions with small clathrin spots. Interestingly, all these endocytic-related structures are polarized within the T lymphocyte uropod.

Our experiments with pharmacological inhibitors indicate that actin cytoskeleton is required for CME in T lymphocytes. This is consistent with data from several other mammalian cell types, in which actin functions at multiple stages of CME (Fujimoto et al., 2000; Yarar et al., 2005). We found that myosin II inactivation produces a partial inhibition and a modest kinetic delay in Tfn uptake. Notably, short-term myosin II blockade, which is not sufficient to abolish morphological polarity, is enough to fully unpolarize CME. Hence, myosin II contractility, in addition to supplying the physical force for uropod retraction (Eddy et al., 2000; Mitchison and Cramer, 1996), may contribute to asymmetric membrane internalization by localizing CME activity to this structure. Consistent with myosin II blockade, inhibition of RhoA/ROCK produces the same redistribution of CME all over the cell surface. Although it is not known how CME and Rho/ROCK/myosin II interact, the rearwards cortical flow could serve as an indirect mechanism to localize CME at the back of the cell. RhoA could also affect CME via its action on the actin cytoskeleton, as a number of actin cytoskeleton regulators bind to endocytic proteins such as clathrin, AP2 and dynamin (Schafer, 2002). Although constitutively active RhoA inhibits CME in non-polarized cells, it stimulates this process in polarized epithelial cells (Lamaze et al., 1996; Leung et al., 1999). Consistent with a common regulation of CME by polarized cells, we observed that RhoA/ROCK inactivation partially inhibits CME. Additionally, activation of RhoA at the posterior pole could increase CME through direct activation of phosphatidylinositol phosphate 5-kinase Iβ, which supplies phosphatidylinositol (4,5)-bisphosphate required to anchor AP2 to the membrane (Gaidarov and Keen, 1999; Padron et al., 2003; Ren et al., 1996). Therefore, RhoA signaling may determine where and how much endocytosis occurs on the surface of a polarized T lymphocyte to regulate membrane retraction during migration.

Cortical flow is the process whereby material in the cell cortex (the outer ∼1-5 μm of the cell) is translocated parallel to the plane of the plasma membrane (Bray and White, 1988). It is widely though that cortical flow is driven by contraction of the cortical actomyosin cytoskeleton and results in the net translocation of cortical F-actin, cortical organelles and cell surface proteins (Benink et al., 2000). Multiple bonds between the membrane and the cytoskeleton may couple both structures during flow. Our data indicate that cargo bound to its receptor flows to large endocytic platforms polarized at the uropod membrane to be primarily internalized there. It could be possible that polarized AP-2 and/or CME could have a role in the generation or maintenance of membrane flow, as suggested by the perturbation of surface-bound Tfn flow by the DIII mutant. However, attempts to directly measure recovery of Tfn fluorescence after bleaching the rear uropod were inconclusive in clathrin-siRNAs or DIII mutant expressing cells (not shown). In addition to AP-2 mediated CME, other plasma membrane clathrin adaptors or non-clathrin mediated endocytic pathways could also be polarized in migrating cells and contribute to generate polarized membrane flow.

The precise function of membrane trafficking in cell migration has been debated for a long time; however, a direct assessment of clathrin function in cell migration has only been analyzed in Dictyostelium. In this regard, clathrin-minus mutants display a reduction in velocity and an increase in their turning ability. To analyze the role of CME during lymphocyte migration, we blunted clathrin expression using siRNA or AP-2 function, using an Eps-15 construct that prevents AP-2 membrane recruitment (Benmerah et al., 1998). Both methods caused a consistent inhibition of basal migration and chemotactic response to the chemokine CXCL12. The importance of chemokine receptor internalization to chemotactic migration is controversial, with data either supporting or ruling out the role of a recycling system in chemotaxis (Richardson et al., 2003; Rose et al., 2004; Yang et al., 1999). Our results reveal a functional role of CME during lymphocyte migration that could affect several distinct aspects of the process: bulk membrane or specific adhesion/chemotactic receptor recycling.

Finally, we propose that polarized endocytic dynamics may provide a mechanism to spatially resolve signaling for guided cell migration. Endocytosis may function as a long-range inhibitor, as it leads to degradation of activated receptors or to recycling to regions of higher signaling (Jekely et al., 2005). Consequently, polarized endocytic activity at the back of a migrating cell may be an effective mechanism to restrict signaling spatially within a cell in order to reinforce polarity. In line with our proposal, polarized endocytosis could be the mechanism that underlies the inhibitory effects of Rho/ROCK and myosin II on sensitivity to attractants at the back of neutrophils (Xu et al., 2003).

Overall, these results suggest the need to revisit the membrane flow model as articulated by Mark Bretscher, who predicted the existence of a posterior endocytic zone to balance the anterior vesicle delivery necessary for membrane growth, and to generate a polarized membrane flow from the anterior to the posterior pole (Bretscher, 1996). In line with this model, we propose that RhoA/ROCK signaling, acting through its effector myosin II, positions the endocytic machinery at the back of a migrating cell; once there, polarized endocytosis cooperates with actomyosin contractility to generate a dynamic cycle of global cortical flow. Finally, polarized endocytic activity may provide the underlying mechanism of spatial cell membrane asymmetries, namely, adhesion assembly/disassembly, sensitivity to attractants and signaling.

The following dyes, fluorescent proteins and antibodies have been used in this study: Alexa 488-, Alexa 633- and Texas Red-conjugated biferric-Transferrin (Molecular Probes); anti-α-adaptin and anti-clathrin heavy chain monoclonal antibodies for immunoprecipitation and immunofluorescence (AP6 and X22, Abcam); anti-β2-adaptin and anti-clathrin heavy chain antibodies (Becton Dickinson), anti-clathrin light chain (CON1, Abcam), and anti-α-tubulin (Sigma) for western blot; anti-α-adaptin polyclonal Ab (M300, Santa Cruz Biotechnology) and anti-Golgin95 (Molecular Probes) for immunofluorescence; Alexa 488-, Cy3- and Texas Red-conjugated goat anti-mouse or rabbit secondary Abs (Jackson Immunoresearch); FM 4-64 plasma membrane red marker (Molecular Probes); DAPI (Sigma). The mouse IgG anti-ICAM-3 (HP2/19), anti-TfnR (FG1/6), and anti-CD45 (D3/9) mAbs were gifts of F. Sánchez-Madrid (Hospital de la Princesa, Madrid, Spain).

Primary T lymphoblasts and the HSB-2 T-cell line were used as spontaneously polarized T-cell models. Peripheral blood mononuclear cells were prepared from single donor leukocyte buffy coats (Centro Regional de Transfusiones, Madrid) and T cells expanded in culture as previously described (Herreros et al., 2000). Primary T cells and HSB-2 T-cell line (ATCC, Rockville, MD) were generally maintained in RPMI 1640 medium (Gibco) supplemented with 10% FCS (Harlam Sera-lab). Most of these cells showed an active leading lamellipodium or pseudopod, and a rear protrusion or uropod in the opposite pole; only ∼30% of cells were unpolarized showing a rounded morphology.

For DNA plasmids and siRNAs transfections, 1-2×10⁷ HSB-2 cells were washed twice in cold RPMI, FCS-free, resuspended in cold Opti-MEM I medium (Gibco) containing 10-20 μg DNA or 300 nM siRNA, and electroporated in an Easyject-Equibio electroporator at 210 V/1200 μF. Imaging was performed 20-40 hours after DNA transfection, or ∼48 hours for siRNAs. Annealed pre-designed siRNAs from Ambion against the clathrin heavy chain were used: ID-107567 (si1), ID-107566 (si2) and ID-107565 (si3). The si1 was not effective and was used as a control, as well as the ID-sc37007 (siC) (Ambion). Interference efficiency was evaluated by immunoblot band densitometry (BioRad).

EGFP-clathrin light chain construct was provided by J. H. Keen (Thomas Jefferson University, Philadelphia, PA) and prepared as described (Gaidarov et al., 1999); EGFP-α-adaptin and the Eps15 mutants DIII-GFP (lacking the EH and the oligomerization domains) and D3Δ2-GFP (lacking EH, oligomerization and AP-2 binding domains, as control) (Benmerah et al., 1998; Rappoport et al., 2003) were provided by A. Benmerah (Cochin Institute, Paris, France); the C3-GFP (TAT-C3 transferase) and the inactive dominant N19RhoA-GFP mutant were provided by F. Sánchez-Madrid.

A confocal scanning inverted AOBS/SP2-microscope (Leica Microsystems, Heidelberg, Germany) provided with a CO₂ and temperature controlled stage was used for time-lapse and static imaging. All images were acquired with a 63X PL-APO NA 1.3 glycerol immersion objective. The theoretical x,y resolution of this lens at Airy-1 and 488 nm excitation length is ∼150 nm and would allow optical z sections of approximately 200 nm depth. Some fluorescent objects under 200 nm were consistently visualized (most Tfn spots), but we do not provide measurements under 200 nm because the apparent size near the resolution limit is a function of both the object size and relative brightness. Isolated clathrin spots (∼200 nm), separated at least 200 nm from other CSs were considered single, small CSs; whereas uninterrupted clathrin arrangements ⩾300-800 nm were scored as large CSs. Different measurements such as areas, lengths or fluorescence intensities in delimited regions of interest (ROIs), were analysed with the LCS 15.37 software (Leica Microsystems). When necessary, images were processed in Adobe Photoshop 5.0.

Assessment of fluorophore co-localization of an image stack was performed with the Leica software that uses a global statistic method to perform intensity correlation analysis. The plots display the intensity distribution and degree of colocalization corresponding to the entire cell, which is shown next to the scatter plot. Co-localizing events in the scatter plot are gated (double positive pixels above the dual threshold) and visualized as a white overlay on the green and red merged image. When co-localization is summarized on single images projected from multiple time/space intervals, the mask coincided with the sum of the co-localizing pixels analysed at each time/space point. An object-based analysis was also preformed to evaluate co-localization in Fig. 3, drawing a vector through a two-dimensional image and plotting the fluorescence intensities for the green and red channel against the length of the vector.

For static immunofluorescence studies, 5×10⁵ cells were incubated for 30 minutes at 37°C and 5% CO₂ on coverslips coated with poly-L-lysine (PLL), 10 μg/ml ICAM-1-Fc, or 10-15 μg/ml fibronectin (FN) and fixed with 4% paraformaldehyde for 30 minutes, permeabilized when necessary with 0.2% Triton X-100 in TBS for 2-5 minutes and blocked in TNB for 15 minutes. Cells were incubated with the corresponding primary antibodies for 1 hour at 37°C, washed in PBS and incubated with labeled secondary antibodies for 30 minutes at 37°C. Cells were finally washed in PBS and water, and mounted with fluorescence mounting medium (DAKO).

In vivo imaging of migrating T cells were assayed on coverslips coated with 2.5 μg/ml ICAM-1-Fc and blocked with 1% BSA for 30 minutes. Occasionally, pinhole size was increased up to 3 Airy units (∼1 μm depth) and 3-5 section stacks were acquired at 20-30 second intervals to visualize wider cell volumes, then maximum projections of time and/or spatial sections were used to integrate events occurring along time in single images.

For photobleaching assays (not formal FRAP), cells were allowed to adhere to FN (20 μg/ml) coated coverslips, incubated on ice for 5 minutes with 10 μg/ml Tfn-A488 or anti-CD45-FITC (Becton Dickinson) as control, and washed twice with cold RPMI. Then, cells were incubated at 37°C and immediately bleached at their apparent rear or front poles. This lower temperature was necessary to slow down CME and get time enough to select the cell and bleach its membrane. Regions of interest (ROIs) with identical size and shape were analyzed for recovery; data were background subtracted, bleaching corrected and normalized before plotting. Besides, we estimated the recovery ratio at each bleached pole as the mean intensity of the bleached membrane divided by the intensity of the opposite unbleached membrane.

T cells (3×10⁵) were pre-incubated in RPMI plus 10% FCS at 37°C and 5% CO₂ for 30 minutes on ICAM-1-Fc or PLL-coated coverslips before the inhibitors addition. Different incubation times were checked: C3-exonenzyme (50 μg/ml, 15 hours; Calbiochem), Y-27632 (20 μM, 2 hours; Calbiochem), blebbistatin (30 μM, 10-30 minutes and 1 hour; Calbiochem), cytochalasin D (2.5 μg/ml, 1-2 hours; Sigma), latrunculine B (0.6 μg/ml, 3-20 minutes; Sigma) and colchicine (1 μg/ml, 1-2 hours; Sigma). After the pre-incubations, 5-10 μg/ml Tfn-A488 was added and cells allowed to endocytose for 5 or 10 minutes before fixation. Rescue assays were performed washing the cells with fresh RPMI medium for 5-10 minutes before Tfn addition.

For uptake assessment the cells were generally pulsed for 1 minute with Tfn, washed twice in RPMI medium, and incubated for up to 20 minutes at 37°C. After the Tfn uptake assay, cells were fixed with paraformaldehyde and surface stained for TfnR. The mean intensity was measured in cytoplasmic ROIs depicted around the endosome compartment using the Leica software. Mean intensity (ROI)=ΣI_i/N_Pixel (I_i, intensity of each pixel, N_Pixel, number of pixels with intensities above the background threshold).

For Rho/ROCK/myosin II-disrupted cells, where endosomes appear distributed all over the cell, internalized Tfn was analyzed in dual channel colocalization plots (Tfn/TfnR). The statistic analysis was performed for the gated `single green events', which correspond to internalized Tfn, excluding `double positive events', which correspond to surface Tfn (supplementary material Fig. S3). The mean and total [mean intensity (au) × masked area (μm²)] intensities were measured in dot plots corresponding to z maximum projections.

5×10⁶ cells were incubated for 2 hours with 20 μM Y-27632 or for 1 hour with 30 μM blebbistatin before lysis in digitonin buffer (1% digitonin w/v; 20 mM triethanolamine; 300 mM NaCl; 2 mM EDTA; 20% glycerol v/v; 10 μM Na₃VO₄; 1 mM PMSF; 1 μg/ml protease inhibitors cocktail). Lysates were pre-cleared with G-protein-Sepharose (Amersham) and incubated with 5 μg/ml X22 or AP6 antibodies for ∼14 hours at 4°C, and then 2 hours with G-protein-Sepharose before precipitation. Immuno-precipitated proteins were revealed by western blot and evaluated by densitometry (BioRad).

Chemotaxis assays were performed using polycarbonate transwell membranes of 5 μm pore size (Costar, Corning, NY). For Eps15 mutants, 1×10⁶ cells were added in 100 μl RPMI supplemented with 10% FCS to the upper chamber, and allowed to migrate for 45 minutes at 37°C in response to CXCL12 (100 ng/ml). Counts of migrated and non-migrated cells were calculated by quantitative fluorescence cytometry (TruCOUNT, FACScalibur, Becton Dickinson), as described (de la Rosa et al., 2003). Migrated and non-migrated cells were analyzed in function of their GFP intensity, and grouped as R1, R2, R3 and R4 populations (R0 corresponded to not transfected). The effect of cell type (mutant vs controls) and transfection level (R0, R1, R2, R3, R4) on the migration ratio (migrated/non-migrated cells) was evaluated by a two-way ANOVA analysis. A similar approach was designed for HSB-2 cells transfected with siRNAs against the clathrin-HC. 1×10⁶ cells were allowed to migrate for 45 minutes at 37°C for 1 hour without or with CXCL12 (100 ng/ml). Differences in the migration rates were evaluated by one-way ANOVA followed by the Bonferroni test for selected pairs of data. The level of significance was P<0.05 in all cases (Statgraphics Plus software).

We thank J. H. Keen, A. Benmerah and F. Sánchez-Madrid for constructs or antibodies; A. Benmerah, J. L. Rodríguez-Fernández and J. M. Serrador for excellent discussions; S. Sánchez-Ramón and E. Obregón for critical reading of the manuscript; I. Sánchez-Ramos for assistance with biostatistics; and J. Villarejo and I. Treviño for expert technical assistance. This work was supported by grants SAF2006-08615 and GEN2003-20649-C06-04 from Ministerio de Educatión to P.S.-M., and fellowships from the Fondo de Investigatión Sanitaria to R.S. and from the Ministerio de Educatión to L.S.-M.