Liprin-α1, ERC1 and LL5 define polarized and dynamic structures that are implicated in cell migration

Cell migration is fundamental for development and tumor metastasis (Sahai, 2007; Vicente-Manzanares et al., 2009), and requires coordination of adhesion, cytoskeletal reorganization and membrane traffic (de Curtis and Meldolesi, 2012; Gauthier et al., 2012). An important question is how these processes are coordinated. Here, we investigated the mechanisms driving protrusion at the edge of cells migrating on or through extracellular matrix (ECM). We recently found that the adaptor protein liprin-α1 is required for efficient cell migration (Asperti et al., 2009). The liprin family includes four liprin-α and two liprin-β proteins that might form homo- and hetero-dimers, and might interact directly with several partners (de Curtis, 2011). The function of liprin-α proteins has been explored mainly in neurons, where they play a central role in regulating synaptic vesicle exocytosis (Stryker and Johnson, 2007).

Liprin-α1 is the most widely expressed member of the liprin-α proteins. Its polypeptide is highly expressed in human breast cancer and other types of human tumors (Rezaul et al., 2010; Astro et al., 2011). Moreover, liprin-α1 is needed for breast cancer cell motility and invasion in vitro (Astro et al., 2011), and is an important regulator of focal adhesion turnover and lamellipodial dynamics. To further dissect the role of liprin-α1 in cell motility, we have looked for relevant liprin-α1-interacting partners. Here, we have identified the adaptor proteins ERC1 (ELKS/Rab6-interacting/CAST family member 1) (Hida and Ohtsuka, 2010; Südhof, 2012) and LL5 (Paranavitane et al., 2003; Kishi et al., 2005) as important co-players, with liprin-α1, in the regulation of tumor cell migration and invasion. The ERC family (Nakata et al., 1999; Monier et al., 2002; Deguchi-Tawarada et al., 2004) includes ERC2, and two ERC1 isoforms, the brain-specific ERC1b and the widely expressed ERC1a (ELKSε) (Wang et al., 2002). In neurons, ERC proteins cooperate with liprin-α in the assembly of functional presynaptic sites (Dai et al., 2006). ERC1a interacts with LL5β (encoded by PHLDB2), a member of the LL5 family of adaptor proteins with a pleckstrin homology (PH) domain, which also includes LL5α (encoded by PHLDB1) and different splice variants. LL5β is part of a stable cortical platform at the periphery of the cell (Paranavitane et al., 2003; Lansbergen et al., 2006; Hotta et al., 2010; Takabayashi et al., 2010). Here, we found that ERC1a and LL5 proteins are needed, together with liprin-α1, to stabilize the edge of migrating cells and for the internalization of active integrins. Liprin-α1, ERC1a and LL5 colocalize near the protruding cell front, where they define new dynamic and highly polarized cytoplasmic structures that are specific to migrating cells. The PH domain of LL5 is required for the integrity of these assemblies and for efficient migration.

Liprin-α1, ERC1a and LL5 proteins are co-expressed in invasive human breast cancer MDA-231 cells (Fig. 1A,B), and can be efficiently silenced with small interfering RNA (siRNA) (Fig. 1B,C; supplementary material Fig. S1A,B). In most experiments, we silenced both LL5α and LL5β to prevent compensatory effects. A complex between overexpressed liprin-α and ERC1, and endogenous ERC2–liprin-α complexes from synaptosomes have been identified (Ko et al., 2003). Recently a complex including liprin-α1, ERC and LL5 was identified at the cortex of non-motile HeLa cells (van der Vaart et al., 2013). Here, complexes containing co-transfected liprin-α1, ERC1a and LL5β were revealed (supplementary material Fig. S1C–E), suggesting that hetero-trimeric complexes might form in cells. However, we could not detect endogenous complexes. Given the dynamic nature of the compartment identified by these proteins, one possibility is that the interaction among the endogenous proteins undergoes a tight temporal regulation preventing the detectable accumulation of the complexes.

As previously described for liprin-α1, depletion of ERC1a or LL5 inhibited MDA-231 cell spreading and migration on fibronectin (supplementary material Fig. S1F–H), as well as haptotactic migration and Matrigel invasion (Fig. 1D,E). The specificity of the effects of siRNAs was supported by reproducing similar effects on migration by a second set of siRNAs (supplementary material Fig. S1I). We isolated independent clones of MDA-231 cells overexpressing GFP–liprin-α1 (L4, L12 and L17 cell lines) or GFP (C9) (supplementary material Fig. S1K). These cells had viability and growth similar to control C9 cells (data not shown), but enhanced invasive abilities that were reversed by silencing endogenous and overexpressed protein (Fig. 1F). Liprin-α1-enhanced invasion was reduced by silencing ERC1a or LL5 (Fig. 1G). These data show a positive role of liprin-α1 in cell invasion and they imply that ERC1a and LL5 proteins have a role in liprin-α1-mediated motility. GFP–liprin-α1-expressing cells moved faster than control cells on two-dimensional (2D) fibronectin-coated substrates (supplementary material Fig. S1L,M). This effect was prevented by depletion of liprin-α1, ERC1a or LL5 (Fig. 1H, supplementary material Fig. S1N), indicating that these proteins are part of a functional network regulating migration.

We analyzed the effects of silencing either protein in a three-dimensional (3D) environment that more closely resembles the mesenchymal matrices met by invading cells in vivo (Cukierman et al., 2001). MDA-231 cells embedded in reconstituted ECMs had an elongated phenotype, with stretches of linear migration suggestive of movement along fibrillar structures (supplementary material Movie 1). Liprin-α1 overexpression enhanced the speed of migration in a 3D environment (Fig. 1I,J). Cells transfected with siRNAs against liprin-α1, ERC1a or LL5 were more rounded than control cells, with less pronounced leading processes and a defect in migration (Fig. 1K,L, supplementary material Movie 1). Although cells depleted of either protein could extend protrusions, these were unstable and the cells were impaired in their ability to move forward. Quantitative analysis confirmed their migration defect, with speed and directionality significantly affected after depletion of liprin-α1, ERC1a or LL5 (Fig. 1M,N).

Silencing of ERC1a or LL5 proteins decreased the stability of lamellipodia (Fig. 2A), as previously observed after liprin-α1 silencing (Astro et al., 2011). Depletion of either protein induced the appearance of lamellipodia more frequently, but these lamellipodia were less persistent (Fig. 2B). The specificity of the effects of the different siRNAs was supported by reproducing similar decreases in the stability of lamellipodia with a second set of specific siRNAs (supplementary material Fig. S1J). In contrast, overexpression of liprin-α1 reduced the frequency and increased the stability of lamellipodia (Fig. 2C). These effects were also observed after the overexpression of ERC1a or LL5β (Fig. 2D). Interestingly, the potentiation of lamellipodial persistence by liprin-α1 overexpression was abolished not only by silencing liprin-α1 but also by silencing ERC1a or LL5 (Fig. 2C). Conversely, the instability of lamellipodia induced by silencing liprin-α1 was partially rescued by ERC1a overexpression (Fig. 2E).

Liprin-α1, ERC1a and LL5 proteins cooperate to regulate cell motility. Previous results have shown that these proteins colocalize in stable patches at the edge of nonmotile HeLa cells (Lansbergen et al., 2006; van der Vaart et al., 2013). Confocal analysis on migrating MDA-231 cells showed that endogenous ERC1a colocalized with liprin-α1 and LL5α near the edge of protrusions (Fig. 3A). ERC1a colocalized with LL5α and LL5β (Fig. 3B), and triple staining for ERC1a, liprin-α1 and LL5β confirmed their colocalization at clusters concentrated near lamellipodia in migrating cells (Fig. 3C). We next explored the dynamic localization of the transfected GFP-tagged proteins in cells migrating on fibronectin. As for the endogenous proteins, the overexpressed proteins were clearly concentrated near protruding lamellipodia (Fig. 3D). Time-lapse images showed a clear relationship between the presence of a protruding lamellipodium and the accumulation of the protein behind it (supplementary material Movie 2). Virtually all actively protruding lamellipodia analyzed (n = 19–22 per construct) showed accumulation of liprin-α1, ERC1a or LL5 proteins.

The dynamic distribution in migrating cells indicates that the three proteins represent markers of highly dynamic functional cytoplasmic structures with a specific and polarized localization near the edge of active protrusions. Analysis of single lamellipodia showed that the concentration of ERC1a becomes evident soon after the protrusive activity starts, and that it is specifically localized behind the protruding lamellipodium. ERC1a-positive clusters disappear rapidly as the lamellipodium halts and starts to retract (Fig. 3E,F). Co-expressed mCherry- and GFP-tagged pairs of the components of the complex partially colocalized near the front of migrating cells (supplementary material Fig. S2).

Interestingly, time-lapse images at higher time resolution showed small motile GFP–ERC1a-positive vesicle-like structures appearing near the protruding cell edge, and moving centripetally towards larger dynamic ERC1a-positive structures (supplementary material Movies 3, 4; Fig. 4A–C). The characteristics of these structures suggest that they might be endocytic vesicles. Several exocytic and endocytic markers were tested to identify the dynamic peripheral structures marked by liprin-α1, ERC1a and LL5. A partial colocalization of ERC1a-positive clusters with caveolin, but not with clathrin, was detected in some analyzed cells (Fig. 4D). No colocalization was detected with any other endocytic and Golgi-associated markers tested, nor with functional markers for clathrin-dependent endocytosis, fluid-phase uptake, or clathrin- and dynamin-independent carriers (CLICs) (supplementary material Fig. S3).

Accumulation of endogenous ERC1a near the leading edge of migrating cells was proximal to, but distinct from, peripheral paxillin-positive focal adhesions (Fig. 5A). The dynamic structures labeled in vivo by GFP–ERC1a (Fig. 5B) and GFP–LL5β (supplementary material Movie 5) were often very close to mCherry–zyxin-positive focal adhesions (Beckerle, 1997), which turn over rapidly at the front of migrating cells. These findings show a tight spatial relationship between the ERC1a- and LL5β-positive structures and dynamic focal adhesions at the protruding front.

There is growing evidence that focal adhesion turnover requires endocytosis and recycling of integrins to and from the cell surface, and that this endocytic cycle is relevant to cell locomotion (Caswell et al., 2009; Valdembri and Serini, 2012). We tested the hypothesis that the liprin-α1, ERC1a and LL5 proteins are involved in integrin endocytosis. Knockdown of either protein did not decrease the density of total β1 integrin at the cell surface detected with the monoclonal antibody (mAb) TS2/16 (Hemler et al., 1984) (Fig. 5C,D); rather, it increased the amount of internalized total β1 integrin after correction for the projected cell area (Fig. 5E,F). We next examined the internalization of active β1 integrin, which accumulates in endocytic compartments (Fig. 6A). Interestingly, depletion of ERC1a, liprin-α1, or LL5 reduced the proportion of internalized active β1 integrin (detected by mAb 9EG7; Lenter et al., 1993) (Fig. 6B,C) by 59.6% (±6.3, s.e.m.) after ERC1a silencing, by 64.8% (±6.5, s.e.m.) after liprin-α1 silencing, and by 40.5% (±8.0, s.e.m.) after LL5 silencing. The decreased internalization of active integrins was not due to a decrease in the density of active integrin β1 at the cell surface, which was indeed increased following ERC1a silencing (Fig. 6D,E). In support of a role of the three proteins in the regulation of integrin traffic, we used an alternative approach that allows the quantification of integrin endocytosis at short times of internalization (Roberts et al., 2001). Biotin-labeled integrins were internalized for 30 min at 37°C before immunoprecipitation with anti-integrin-β1 antibodies (Fig. 6F). Knockdown of either protein (Fig. 6G) had a negative effect on integrin internalization, with significant decreases measured after liprin-α1 or ERC1a depletion (Fig. 6H).

To test for the specificity of the effects observed on integrin internalization, we looked at the effects of ERC1a depletion on other functional endocytic markers. We used fluorescently labeled transferrin to assess clathrin-mediated endocytosis (supplementary material Fig. S3D,E), fluorescently labeled dextran as a marker of fluid-phase endocytosis (supplementary material Fig. S3F,G), and cholera toxin B as a marker of CLICs (supplementary material Fig. S3H,I). No effects were observed on transferrin and cholera toxin B internalization after silencing ERC1a compared to control cells. ERC1a-depleted cells internalized less dextran, but when the internalization was normalized for the reduced projected cell area induced by ERC1a silencing, no differences were found compared to control cells (supplementary material Fig. S3G).

Given the inhibitory effects observed on cell speed upon depletion of either liprin, ERC or LL5 proteins, these results suggest that liprin-α1, ERC1a and LL5 proteins might promote cell migration by influencing the internalization of active integrins, which are required for adhesion turnover at the cell front. In this direction, we assessed the effects of liprin-α1 depletion on the dynamics of the focal-adhesion-positive edge of migrating cells. We found that silencing of liprin-α1 significantly increased the fraction of projected cell area containing mCherry–zyxin-positive focal adhesions (Fig. 6I,J). Moreover time-lapse on living cells co-transfected with mCherry–zyxin and siRNA against liprin-α1 showed a reduction in the dynamics of the mCherry–zyxin-positive focal adhesion area with respect to cells co-transfected with mCherry–zyxin and control siRNA (Fig. 6K,L).

Liprin-α1, ERC1a and LL5β colocalize in a highly polarized and dynamic manner near the protruding edge of migrating cells (Figs 3, 4). We tested whether the accumulation of each protein at sites of protrusion was dependent on the other two proteins. We looked at the subcellular distribution of each endogenous protein after depletion of the other two proteins, and evaluated the accumulation of clusters positive for each protein at protrusions, which were identified as detailed in the Materials and Methods. Analysis by immunoblotting confirmed that single or combined depletion by siRNAs was efficient under all the experimental conditions tested (supplementary material Fig. S4). The subcellular distribution of endogenous ERC1a was strongly affected by depletion of either liprin-α1 or LL5 proteins (the percentage of protrusions per cell containing ERC1a-positive clusters was reduced by 60.1% and 92.2% compared with control, respectively) (Fig. 7A,B). By contrast, the accumulation of LL5β-positive clusters at protrusions was strongly affected by depletion of liprin-α1 (the percentage of protrusions per cell containing LL5β-positive clusters was reduced by 70.6%), but not by ERC1a silencing (Fig. 7C,D). Finally, the accumulation of liprin-α1-positive clusters at protrusions was strongly affected by LL5 depletion but only mildly affected by ERC1a silencing (the percentage of protrusions per cell containing liprin-α1-positive clusters was reduced by 86.9% and 21.9% compared with control, respectively) (Fig. 7E,F). Therefore, although liprin-α1 and LL5β are dependent on each other for their accumulation at protrusions, ERC1a does not play a major role in their localization at these sites. By contrast, the localization of ERC1a depends on both liprin-α1 and LL5 proteins. Depletion of each protein or of pairs of proteins had no or only very mild effects on the number of protrusions per cell (Fig. 7G). This is not surprising because the analysis from time-lapse images showed that lamellipodia could still form in cells silenced for the endogenous proteins, but that they were less stable (Fig. 2A,B). Taken together, these results suggest a functional link between the colocalization of liprin-α1, ERC1a and LL5 at the dynamic polarized structures and the effects of these proteins in stabilizing the protrusive activity during migration. In this regard, depletion of liprin-α1 in cells overexpressing GFP–ERC1a caused the loss of GFP–ERC1a-positive clusters from the cell periphery, as well as the formation of less stable lamellipodia (Fig. 7H), as observed following liprin-α1 depletion in cells with endogenous levels of ERC1a (Fig. 2A,B). A significant increase in the number of focal adhesions was observed in cells overexpressing GFP–ERC1a (Fig. 7I,J). Moreover, ERC1a overexpression induced a faster turnover of the focal adhesions. Both effects required endogenous liprin-α1 because they were inhibited by liprin-α1 silencing (Fig. 7K).

LL5 proteins include a C-terminal PH domain that recognizes phosphatidylinositol-3,4,5-triphosphate (PIP₃), and is required for the subcellular localization of LL5β at the cell cortex (Paranavitane et al., 2003). We investigated the requirement of the PH domain of LL5β for the localization and function of the trimeric complex. Under conditions of low phosphoinositide 3-kinase (PI3K) activity (serum starvation and wortmannin treatment) or upon expression of LL5β with a mutation in the PH domain that prevents binding to PIP3 (LL5β^M), LL5β predominantly localizes to punctate vesicular structures that are dispersed throughout the cytoplasm (Paranavitane et al., 2007; Takabayashi et al., 2010). The overexpression of the PH mutant LL5β^M (Fig. 8A) induced the expected accumulation of the mutant at cytoplasmic vesicles, but did not interfere with the accumulation of endogenous liprin-α1 at the periphery of cells with normal levels of endogenous LL5 proteins (Fig. 8B,D). In addition, the dispersion of liprin-α1 from the peripheral protrusions of LL5-silenced cells (Fig. 7E,F) was rescued by the expression of siRNA-resistant GFP-tagged mouse wild-type LL5β, but not by the expression of the LL5β^M PH mutant (Fig. 8C,D): endogenous liprin-α1 remained mainly diffuse localized throughout the cytoplasm and did not colocalize with the GFP–LL5β^M-positive dots induced by the mutant (Fig. 8C). The expression of GFP–LL5β^M either in cells with endogenous LL5 (Fig. 8E) or after silencing LL5 (Fig. 8F) caused the recruitment of endogenous ERC1a to the GFP–LL5β^M-positive dots, probably owing to the interaction of ERC1a with LL5β^M (supplementary material Fig. S1D). These results indicate that the PH domain of LL5β is required for the accumulation of the trimeric complex near the protrusions. However, the presence of the PH domain is not sufficient for the recruitment of the complex, because our data in Fig. 7 indicate the additional requirement of liprin-α1 for the accumulation of the complex at the protrusions.

The silencing of LL5 proteins causes a defect in migration (Fig. 1) and in lamellipodial dynamics (Fig. 2). The defect in migration was rescued by the expression of the GFP-tagged wild-type LL5β protein, but not by the GFP–LL5β^M PH mutant (Fig. 8G). Analysis of lamellipodial dynamics demonstrated that the defect induced by LL5 silencing was fully rescued by the wild-type GFP–LL5β, and only partially rescued by GFP–LL5β^M, given that a significant defect in lamellipodial persistence was still observed in cells expressing the PH mutant (Fig. 8H,I). Taken together, these results show that the PH domain of LL5β is required for the proper localization of liprin-α1 and ERC1 at the cell edge, and support the hypothesis that there is a functional link between the subcellular accumulation of the three proteins to the protrusions and their requirement for the stability of the lamellipodia and efficient migration.

Our results show that liprin-α1, ERC1a and LL5 proteins cooperate to drive efficient migration on and through ECM. They do so by stabilizing the lamellipodia at the front of migrating cells, where the three proteins partially colocalize in newly identified dynamic cytoplasmic structures that are highly polarized and specifically concentrated near the active protrusions. The effects of the silencing of any of the three proteins on the stability of the lamellipodia and their subcellular colocalization indicate that they are part of the same functional network. The negative effects of the silencing of any of the proteins on the lamellipodial stability in 2D and 3D environments might explain the consistent defects observed in the migration and invasion of the silenced cells.

Liprin-α1, ERC1/ELKS and LL5β have previously been shown to colocalize at cortical platforms at the periphery of non-motile cells, where these proteins are required to stabilize microtubules through CLASP proteins. These sites displayed no significant motility, behaving as stable and relatively immobile structures at the periphery of HeLa cells (Lansbergen et al., 2006; van der Vaart et al., 2013). In contrast, the structures identified by the three proteins in actively migrating MDA-231 are highly dynamic and concentrated to restricted areas of the cell corresponding to active protrusions, where they partially colocalize with caveolin.

ERC proteins have been implicated in membrane trafficking: ERC2 and liprin-α regulate presynaptic vesicle traffic in neurons (Dai et al., 2006); in non-neuronal cells ERC1 participates in the docking and/or fusion of Rab6-positive vesicles at the cell cortex and, together with liprin-α, in secretion in mast cells (Grigoriev et al., 2007; Nomura et al., 2011). Here, we observed a strong and specific inhibition of the internalization of active β1 integrin after depletion of ERC1a, liprin-α1 or LL5 proteins, whereas other endocytic pathways were not affected. Based on these results and on the identification of ERC1a-positive dynamic vesicle-like structures by video microscopy, we speculate that the identified cytoplasmic structures represent a functional state of an endocytic compartment characterized by the recruitment of ERC, LL5 and liprin-α1 on the cytoplasmic side of intracellular membranes.

The localization of LL5β at the cell membrane is dependent on PIP₃ (Paranavitane et al., 2003). Mutation of the LL5β PH domain perturbs the accumulation of the other proteins of the complex at the cell periphery, and negatively affects lamellipodial stability and cell motility. Based on these results and on the observed negative effects of the depletion of each component on the localization of the remaining proteins, and on the tight association of liprin-α1 to the cytoplasmic side of the plasma membrane (Asperti et al., 2009; Asperti et al., 2010), we propose a model in which liprin-α1 and LL5β are recruited with ERC1a to the cytoplasmic side of the plasma membrane and, eventually, to the cytoplasmic side of endocytic vesicles during cell migration (Fig. 8J). This mechanism might be relevant to the turnover of adhesions observed at the front of highly motile cells.

In this regard, in actively migrating cells, the three proteins colocalize dynamically, close to the centripetal side of the focal adhesions at the protruding edge, more distal from the plasma membrane. The rapid dynamics of focal adhesion at these sites is expected to require fast active integrin turnover. There is growing evidence that internalized integrins can be rapidly recycled at the front of migrating cells: localized endocytic and exocytic traffic of integrins can spatially restrict receptor distribution to promote the turnover of adhesions and signaling molecules at the cell front and to support protrusion (Caswell et al., 2009). It is known that β1 integrin can be internalized (Pellinen et al., 2008; Shi and Sottile, 2008) in both inactive and ligand-bound active conformations (Ng et al., 1999; Valdembri et al., 2009; Dozynkiewicz et al., 2012), and that the two pools might be recycled through distinct routes (Arjonen et al., 2012). The negative effect of the lack of liprin-α1, ERC1a or LL5 on the specific internalization of active β1 integrin supports the hypothesis that the dynamic structures identified in this study are directly or indirectly involved in the internalization of integrins, thus contributing to the turnover of focal adhesions at the front of migrating cells. Although a clear accumulation of β1 integrin in the peripheral ERC1a-positive structures could not be detected (our unpublished data), a rapid and transient association of the internalized receptors with the proposed ERC1a-positive structures cannot be excluded. Interestingly, in polarized epithelial cells, LL5 proteins colocalize with laminin-5 and its integrin receptors α3β1 and α6β4 at the basal side, possibly owing to indirect interaction of LL5 with the cytoplasmic tails of the α-integrin subunits (Hotta et al., 2010).

In conclusion, we would like to propose that the highly polarized and dynamic structures specified by liprin-α1, ERC1a and LL5 proteins identify a new mechanism that is important for lamellipodia stabilization and adhesion turnover, both of which are required for efficient protrusion during migration.

Plasmids for FLAG–liprin-α1 and GFP–liprin-α1 (human), GFP–LL5α (human), GFP–LL5β (mouse), GFP–ERC1a (full-length murine ELKSε), mCherry–zyxin, and Mito-RFP were as described previously (Totaro et al., 2007; Asperti et al., 2009; Hotta et al., 2010; Kishi et al., 2005; Inoue et al., 2006; Colombelli et al., 2009). GFP–LL5β^M was generated by PCR with primers 5′-GGTGGGAAAATTAAAACGTGGAAAGCAGCCTGGTTTGTTTTTGACCGGAATAAG-3′ and 5′-CTTATTCCGGTCAAAAACAAACCAGGCTGCTTTCCACGTTTTAATTTTCCCACC-3′ to give two mutations (K1211A, R1212A) in the PH domain of murine LL5β (accession number NP_001239371.1). The mCherry–liprin-α1, mCherry–ERC1a and mCherry-C1–LL5β were generated from plasmids for the corresponding GFP-tagged proteins. The pEGFP-Liprin-α1^SR plasmid was obtained by site-direct mutagenesis on pEGFP-Liprin-α1 (primers: 5′-TTAACCCAGGGGAAGTTACACGAAGTGGGTCACGAAAGAGATTCCTTGCAGAGACAGCTC-3′ and 5′-GAGCTGTCTCTGCAAGGAATC TCTTTCGTGACCCACTTCGTGTAACTTCCCCTGGGTTAA-3′).

The polyclonal antibody (pAb) against liprin-α1 was as described previously (Asperti et al., 2009); the commercial mAb against liprin-α1 was from Proteintech. The mAb for both LL5α and LL5β (clone IH12), the hamster mAb against LL5α (223), and pAb against mannose 6-phosphate receptor were as described previously (Ludwig et al., 1991; Hotta et al., 2010; Kishi et al., 2005). Other mAbs were specific for: Golgi 58K, ERC1a (ELKS-30), Lamp1 (H4A3) from Abcam; EEA1, 9EG7 specific for active β1 integrins, paxillin (349) and GM130 (35) from BD Biosciences; Rab6 (D37C7) from Cell Signaling; TS2/16, recognizing the extracellular portion of β1 integrin, from American Type Culture Collection; tubulin-α (Clone DM 1A) from Sigma; clathrin heavy chain (TD.1) from Santa Cruz Biotechnology. The β1 integrin pAb used for immunoprecipitation was as described previously (Tomaselli et al., 1988). Commercial pAbs specific for GFP (Life Technologies), fibronectin and LL5β (Sigma); caveolin-1 (N-20), giantin (H-49), paxillin (H-114) (Santa Cruz Biotechnology); GFP (chicken pAb; Abcam). FITC- and TRITC-Phalloidin were from Sigma. Alexa-Fluor-conjugated secondary antibodies, dextran–Alexa-Fluor-568, transferrin– Alexa-Fluor-647, CtxB (cholera toxin B)–Alexa-Fluor-555 and horseradish peroxidase (HRP)-conjugated streptavidin were from Life Technologies. EZ-link cleavable sulfo-NHS-SS-biotin (#21331) was from Thermo Scientific; MesNa (sodium 2-mercaptoethanesulfonate, #63705) was from Fluka, and iodoacetamide from Sigma.

Cells were transfected with Lipofectamine-2000^® (Life Technologies) with 1–2 µg of plasmid and/or 50–100 nM of siRNAs, and used 24–48 h later. siRNAs (Life Technologies) for luciferase, liprin-α1, ERC1a, LL5β and LL5α were as described previously (Asperti et al., 2009; Lansbergen et al., 2006; Ducut Sigala et al., 2004; Hotta et al., 2010; Astro et al., 2011). The region targeted by siRNA for human LL5β includes two mismatches with respect to the mouse cDNA sequence, making this latter resistant to silencing. Second siRNAs for human LL5α (5′-GATGGGTTAGCCACCCGTA-3′) and LL5β (5′-GGATCTACCTCATAGCGTA-3′) were from Qiagen.

Cell lines stably expressing either GFP–liprin-α1 or GFP were obtained after transfection of MDA-231 cells with 2 µg of either GFP–liprin-α1 or GFP plasmid, respectively. Clones resistant to 1 mg/ml of G418 (Merck Millipore) were expanded from single cells isolated by limiting dilution.

Immunoprecipitates from cell lysates incubated with Protein-A–Sepharose beads (Amersham Biosciences) conjugated to antibodies, were separated by SDS-PAGE for immunoblotting.

Cell viability was assessed by a colorimetric assay based on the reduction of MTT [3-(4,5-dimethylthiazol-2yl)-2,5-diphenyl-tetrazolium bromide]. For cell proliferation, 2000 cells were seeded in 24-well plates and their growth was evaluated by counting in Burker chamber every 24 h.

Haptotactic migration, cell spreading, and invasion assays were as described previously (Astro et al., 2011). Data were normalized to control and presented as means±s.e.m. For random migration, transfected cells seeded for 2 h at 37°C on 2.5 µg/ml fibronectin (50,000 cells per 3.5-cm diameter well) were imaged for 5 h at 6-min intervals with an Axiovert microscope (Zeiss) equipped with an Orca II CCD camera (Hamamatsu), and analyzed with ImageJ (NIH). For lamellipodial dynamics, frames were analyzed with ImageJ by counting the number of lamellipodia forming during 1 or 5 h. Persistence of protrusions was evaluated as the time elapsed between the formation of a lamellipodium and its disassembly. Statistically significant differences were analyzed Student's t-test.

3D matrices were prepared as described previously (Cukierman et al., 2001; Cukierman, 2005). Transfected MDA-231 cells were seeded on matrices for 6 h before microscopy. Frames were acquired every 7 min for 8 h, and analyzed with ImageJ.

MDA-231 cells (10⁵ cells) in fibronectin-coated (2.5 µg/ml) coverlips were transfected after 1 day, and used for video imaging or immunofluorescence 24–48 h later (Astro et al., 2011). Live imaging was performed on Inverted Eclipse TE2000-E microscope (Nikon) with a spinning-disc head (CSU-10, Yokogawa) equipped with a Cool SNAP HQ2 CCD camera (Photometrics). Pictures were at 1 frame/min or 0.5 frames/s. Images of fixed cells were acquired with a confocal Ultraview ERS system from PerkinElmer.

The quantification of the accumulation of clusters positive for liprin-α1, ERC1a and/or LL5 proteins at protrusions was performed as follows: we considered as protrusions the cellular extensions that showed a clear F-actin-positive edge. We counted the number of protrusions with protein-positive clusters and the number of total protrusions in each cell. We calculated the ratio between these two values (number of protrusions with protein-positive clusters/number of total protrusions). Average ratios from several cells for each experimental condition are expressed as percentages; the numbers of cells considered for each condition are indicated in the respective figure legends.

Transfected MDA-231 cells on fibronectin-coated (2.5 µg/ml) coverslips were incubated overnight at 37°C, under 5% CO₂ in growth medium with either 2 µg/ml TS2/16 mAb, or 1.25 µg/ml 9EG7 mAb (Margadant et al., 2012). Cell-surface-bound antibodies were stripped with 0.5 M NaCl and 0.5% acetic acid, pH 2. After fixation and permeabilization with 0.1% saponin, cells were immunostained with Alexa-Fluor-conjugated secondary antibodies. Quantification of internalized integrin–antibody complexes was performed with ImageJ. For quantitative analysis, exposure time, gain and offset were adjusted to avoid saturated pixels. Images were acquired at the same exposure and identically thresholded integrin images were used to measure the area occupied by internalized integrins (positive for 9EG7 or TS2/16). The projected cell area was measured on thresholded GFP images. Values obtained are represented as fold increase compared to control cells (siRNA against luciferase) of: (1) the projected cell area; (2) the area occupied by internalized integrins; and (3) the fraction of the cell area occupied by internalized integrins. Data were from two or three experiments in which at least 60 cells were analyzed.

We used a modification of the biochemical method described previously to quantify integrin internalization at shorter times, based on the immunoprecipitation of biotinylated integrins (Roberts et al., 2001). MDA-231 cells transfected with siRNAs were cooled on ice, rinsed twice with PBS and exposed for 30 min at 4°C to 0.5 mg/ml EZ-link cleavable sulfo-NHS-SS-biotin (Thermo Scientific) in Hank's balanced salt solution (Life Technologies). After removing unbound biotin with cold medium, the cells were either lysed immediately to identify the surface receptor pool, or incubated for 30 min at 37°C to allow the internalization of integrins. Cells were then washed twice with cold PBS, and the remaining biotin-labeled surface proteins removed by a 45-min incubation at 4°C in 60 mM sodium 2-mercaptoethanesulfonate (MesNA), 100 mM NaCl, 50 mM Tris-HCl pH 8.6, followed by quenching for 15 min on ice with 100 mM iodoacetamide (Sigma). The cells were lysed, and biotinylated integrins were immunoprecipitated with anti-integrin-β1 pAb (Tomaselli et al., 1988) conjugated to protein-A–sepharose beads. As control of protein biotinylation, one dish was lysed after labeling, without treatment with MesNA. Lysates and immmunoprecipitates of internalized or surface biotinylated integrins were detected with HRP-conjugated streptavidin (Life Technologies). Blots were then stripped in acid buffer (0.2 M glycine, 0.1% SDS, 1% Tween-20, pH 2.2) and re-probed with anti-β1-integrin pAb. Biotinylated surface and intracellular β1 integrins were quantified with ImageJ from bands of immunoprecipitates. The relative amount of internalized biotin-labeled integrin were normalized to the corresponding amounts of total biotin-labeled surface integrin (before internalization). Data are expressed as fold increase compared to control cells (siRNA against luciferase) of the integrated density of the bands from blots, from two independent experiments.

MDA-231 cells seeded on 24-mm diameter glass coverlips precoated with 2.5 µg/ml fibronectin (120,000 cells/well) were co-transfected with siRNAs and mCherry–zyxin. Confocal live imaging was performed 48 h later on an inverted spinning-disc Ultraview ERS confocal microscope (PerkinElmer) equipped with a temperature- and CO₂-controlled stage. Images were acquired with a 100× 1.3 NA oil immersion objective (Zeiss) for 30 min at 0.5 frames/min. Identical exposure and laser power were used for all samples. Pictures were taken at the same exposure with a 532-nm laser. Quantitative analysis of focal adhesion density and turnover was performed with ImageJ on thresholded mCherry–zyxin images. The same threshold was used for all conditions within a single experiment. Quantification of focal adhesion density at the cell edge was performed on the first image of each time series. Areas of 20-µm×10-µm were positioned on protrusions to measure the area of peripheral adhesions. Values were expressed as fractions of marginal cell area occupied by adhesions, and were normalized to the values obtained for control cells (siRNA against luciferase). Data were collected from 14–18 active cell protrusions from two or three independent experiments.

For the quantification of adhesions turnover, we used a modification of a published protocol (Grande-García et al., 2007). We measured the changes in mCherry–zyxin-positive area at the periphery of migrating cells on thresholded images collected over 30 min, by applying the operation ‘difference’ of the ‘image calculator’ function of ImageJ. We obtained 15 ‘subtracted’ consecutive frames for each protrusion. The measure of the differences between mCherry–zyxin-positive areas from consecutive frames was defined as ΔFA. From the same 15 subtracted consecutive frames we calculated, in parallel, the changes of the corresponding total area occupied by the protrusion (ΔTOT). The ratio between the ΔFA and ΔTOT values represents a measure of adhesion turnover. These values were normalized to control cells (siRNA Luciferase). Data were from 12 active cell protrusions per conditions, from two or three independent experiments.

For the analysis in cells overexpressing GFP–ERC1a, number and size of mCherry–Zyxin-positive focal adhesions were quantified on thresholded images by the Analize Particles plugin of ImageJ (version 1.48t). Values were represented as fold increase compared with control cells. For quantification of focal adhesion turnover, time of formation and disappearance of each adhesion from protruding areas was monitored, and lifetimes were determined for each condition. Each experiment was repeated at least two times.

Uptake of endocytic markers was performed as described previously (Howes et al., 2010, Fletcher et al., 2012). Briefly, 48 h after transfection cells on fibronectin were incubated for 30 min at 37°C with 1 mg/ml dextran–Alexa-Fluor-568, 20 µg/ml transferrin–Alexa-Fluor-647 or 10 µg/ml CTxB–Alexa-Fluor-555 (Hewlett et al., 1994). After washing and removal of surface-bound transferrin by 0.5 M glycine, pH 2.2, cells were fixed with 3% parafolmaldehyde for confocal analysis.

We thank Joshua R. Sanes (Harvard University, Cambridge, MA) for LL5 mAb and GFP–LL5β, Yuko Mimori-Kiyosue (Kobe Institute, Japan) for LL5α mAb and GFP–LL5α, Yoshimi Takai (Osaka University, Japan) for GFP–ERC1a, Vic Small (Institute of Molecular Biotechnology, Vienna, Austria) for mCherry–zyxin, and Bernard Hoflack (Dresden University of Technology, Dresden, Germany) for the mannose-6-phosphate receptor pAb. We thank Francesca Capriotti, Diletta Tonoli, Cesare Covino (Alembic) for technical help, and Jacopo Meldolesi (San Raffaele Scientific Institute, Milano, Italy) for critical reading of the manuscript.