Flotillin microdomains stabilize cadherins at cell–cell junctions

Cell–cell junction (CCJ) formation and remodelling occur repeatedly throughout development and perturbation of CCJs is associated with cancer cell invasion and metastasis (Christofori, 2003). Cell–cell adhesion is promoted by cadherins, which are transmembrane glycoproteins that form homophilic complexes and clusters at CCJs (Harris and Tepass, 2010). The extracellular region allows homophilic ligation with cadherins present on adjacent cells (Troyanovsky, 2005; Smutny et al., 2010). The intracellular cytoplasmic tail binds to proteins called catenins. The C-terminal end binds directly to β-catenin, which serves as a scaffold to anchor α-catenin. The juxtamembrane domain of the cytoplasmic tail of cadherin binds to p120 catenin, which regulates cadherin stability at CCJs (Reynolds, 2007; Nanes et al., 2012). This defines the core of the cadherin–catenin complexes, but other proteins can associate directly or indirectly with the cytoplasmic tail of cadherins, including signalling molecules, scaffolding proteins and cytoskeletal regulators (Niessen et al., 2011). These interactions determine the dynamics and stability of cadherin adhesive structures. The dynamics of the junctional F-actin cytoskeleton plays a major role in the assembly of cadherin–catenin complexes at CCJs where they act as signalling platforms (Xiao et al., 2005). Our knowledge on cadherin dynamics has increased in recent years, but the mechanisms of cadherin clustering and stabilization at CCJs are still important issues to address.

To identify novel regulators of CCJs homeostasis, we performed a proteomic analysis of the molecules that were immunoprecipitated together with N-cadherin in C2C12 myoblasts and detected flotillin 1 (Reggie-2). Flotillin 1 and flotillin 2 (Reggie-1) are homologous, ubiquitous and highly conserved proteins. They localise at the cytoplasmic face and in specific cholesterol-rich microdomains of the plasma membrane (PM) by acylation, especially myristoylation and palmitoylation, and in intracellular compartments. The N-terminal part of flotillins is a prohibitin homology domain (PHB) that allows association with the PM and interacts directly with F-actin (Langhorst et al., 2007). The C-terminus contains an α-helical region that mediates oligomer formation (Langhorst et al., 2007; Solis et al., 2007). Flotillins define caveolin-independent membrane microdomains involved in endocytosis, cell adhesion and migration (Watabe-Uchida et al., 1998; Stuermer et al., 2001; Stuermer, 2009; Zhu et al., 2012) and play a role in the scaffolding of large complexes that signal across the PM by forming hetero-oligomeric complexes (Langhorst et al., 2007). These properties make flotillins interesting candidates for the regulation of the dynamic association of cadherin complexes at CCJs. Indeed, we previously showed that N-cadherin is stabilized at CCJs through its association with cholesterol-rich membrane domains (Causeret et al., 2005). Flotillins have been detected at cell–cell contact sites (Morrow et al., 2002; Neumann-Giesen et al., 2004; Liu et al., 2005; Glebov et al., 2006) and their potential involvement in CCJs formation was recently suggested (Chartier et al., 2011; Solis et al., 2012); however, their exact role in cadherin-dependent processes and in CCJs homeostasis has not been explored yet.

As we identified flotillins as N-cadherin partners, we first examined whether flotillins also associate with other cadherin types. We demonstrated that flotillin association at CCJs is mediated by all the cadherins we have tested so far and revealed for the first time the existence of cadherin–flotillin complexes at the PM in mesenchymal and epithelial cells. The cadherin–flotillin complexes at CCJs are connected with F-actin bundles and are found in monosialoganglioside GM1-containing membrane microdomains. Knockdown of flotillins inhibited cadherin association with GM1-containing PM microdomains and thus cadherin stabilization at CCJs and the formation of functional CCJs. These results identify flotillins as new regulators of cadherin-mediated CCJs formation.

After having identified flotillins among the proteins that are immunoprecipitated with N-cadherin, we confirmed that flotillins are associated with N-cadherin and also with E-cadherin by immunoprecipitation experiments using PM fractions from mouse C2C12 myoblasts and human breast carcinoma MCF-7 epithelial cells (Fig. 1A,B). The absence of association of the transferrin receptor (TfR, a transmembrane protein localised to the PM and in intracellular vesicles) or of caveolin 1 (a membrane protein) with immunoprecipitated N-cadherin demonstrated the specificity of the association between cadherins and flotillins (Fig. 1C). To investigate whether the N-cadherin–flotillin association required N-cadherin engagement at CCJs, we monitored the effect of homophilic interaction disruption by Ca²⁺ chelation with EGTA on flotillin association with N-cadherin in PM-enriched fractions (Fig. 1D). Flotillin association with N-cadherin was not perturbed by EGTA treatment in both C2C12 myoblasts (Fig. 1D) and MCF-7 cells (unpublished data). These data suggest that the cadherin–flotillin association is constitutive at the PM.

We then examined flotillin 1 and 2 localisation in mesenchymal (C2C12 myoblasts) and epithelial cells (MCF-7, HCT-116 and Caco-2). Flotillin 1 and 2 localisation in C2C12 myoblasts was analysed by expressing mCherry- or HA-tagged flotillins because the available anti-flotillin antibodies were not effective in detecting the endogenous proteins in C2C12 myoblasts and L cells (used later in this study). Analysis of exogenous and endogenous flotillins (Fig. 2 and supplementary material Fig. S1) confirmed their PM and cytoplasmic distribution (Langhorst et al., 2008). Moreover, we observed that, in all tested cell types, exogenous and endogenous flotillins mainly colocalised with cadherins at CCJs (Fig. 2A–C; supplementary material Fig. S1A–C, Movies 1–5) and in lamellipodia (Fig. 2D,E; supplementary material Movie 6), but rarely in intracellular compartments. We also used three-dimensional structured illumination microscopy (3D-SIM), which improves resolution by a factor of two in the x-, y- and z-planes (Gustafsson et al., 2008), to better characterize colocalisation of cadherins and flotillins at CCJs in C2C12 myoblasts and MCF-7 epithelial cells. As a positive control for colocalisation, p120 catenin staining was performed. We showed that N-cadherin and E-cadherin colocalised with flotillin 1 and 2 specifically at the CCJs and not in intracellular compartments, similarly to cadherin colocalisation with p120 catenin (Fig. 3). 3D-SIM also showing that cadherins and flotillins that colocalised at CCJs were associated with F-actin bundles.

In various cell types, newly formed cell–cell contacts can be differentiated from established mature ones (Comunale et al., 2007; Affentranger et al., 2011; Taguchi et al., 2011). In Caco-2 cells, in which these two stages are easily distinguishable, flotillins were barely detectable at new CCJs, whereas they accumulated at mature CCJs (supplementary material Fig. S1C). This was also confirmed in C2C12 myoblasts (supplementary material Movie 6). Moreover, the proportion of cells that establish mature CCJs can be synchronized after restoration of extracellular Ca²⁺ in cells previously treated with EGTA (supplementary material Fig. S1D, recovery). When compared with the steady-state cell populations (supplementary material Fig. S1D, untreated) in which the proportion of mature CCJs is heterogeneous, we observed that cadherin and flotillin colocalisation increased during CCJs formation and stabilization (supplementary material Fig. S1D; Movie 6). These data indicate that colocalisation of cadherins and flotillins at cell–cell contacts increases during CCJ maturation and suggest that flotillins stabilize cadherins at CCJs.

To assess whether accumulation of flotillins at CCJs requires the presence of cadherins, we used L cells that do not express endogenous cadherins. In parental L cells, flotillins did not accumulate at cell–cell contacts. However, when cadherin-N, -E, -P, -R and -11 were expressed in these cells, flotillins accumulated at CCJs where they colocalised with cadherins (Fig. 4A,B and supplementary material Fig. S1E). In agreement, flotillins were co-immunoprecipitated with E-cadherin (Fig. 4C) and N-cadherin (unpublished data) from PM fractions of L cells. To demonstrate that cadherin-driven accumulation of flotillins at CCJs was not simply caused by membrane overlapping, we analysed the distribution of other membrane proteins. First, we investigated the localisation of CD44, an ubiquitously expressed membrane protein involved in cell adhesion and migration (Goodison et al., 1999). Whereas flotillins accumulated with cadherins at CCJs, CD44 did not (Fig. 4A). Line-scan analysis confirmed the specific accumulation of flotillins at CCJs upon expression of cadherin. We also analysed the distribution of caveolin 1, a membrane protein that is not associated with the cadherin–flotillin complex (Fig. 1C) and does not colocalise with flotillins (Watabe-Uchida et al., 1998). Caveolin 1 distribution was comparable in L cells that expressed cadherin and those that did not, and it did not accumulate at CCJs in any case (Fig. 4D). This indicates that flotillin accumulation at CCJs is not simply due to non-specific membrane overlap. Altogether, these findings demonstrate that flotillin accumulation at CCJs is promoted by cadherins. Moreover, flotillins are associated with cadherin complexes mainly at the PM (Fig. 1) and particularly at CCJs, where flotillin–cadherin colocalisation occurs on F-actin bundles.

We then assessed the role of flotillin in cadherin-dependent cell–cell adhesion in mesenchymal (C2C12 myoblasts) and epithelial cells (MCF-7, HCT-116 and Caco-2). We generated, by retroviral infection, stable C2C12 cell lines that express anti-flotillin 1, -flotillin 2 or -luciferase (negative control) shRNAs. A 60% reduction of flotillin 1 protein was observed in C2C12 myoblasts expressing flotillin 1 and flotillin 2 shRNAs (Fig. 5A), as previously reported (Vassilieva et al., 2009). Silencing of flotillin impaired N-cadherin and p120 catenin accumulation at CCJs in comparison with control cells (Fig. 5B). As investigated further in this study (Fig. 8), this defect is due to a reduction in the global levels of cadherin (Fig. 8G–I) correlated with the perturbation of cadherin dynamics at CCJs (Fig. 8C,D). To exclude off-target effects of the anti-flotillin 1 shRNAs, we rescued flotillin 1 expression by transfecting a modified, shRNA-resistant flotillin 1 cDNA (Rflot1). In flotillin 1 shRNA C2C12 myoblasts, mCherry-tagged Rflot1 localised at CCJs where it promoted N-cadherin accumulation and p120 recruitment (Fig. 5C). To check the consequences of flotillin 1 knockdown in epithelial cells, we used a mixture of two previously described anti-flotillin 1 siRNAs (Glebov et al., 2006) (Fig. 5D). In flotillin 1 siRNA MCF-7 cells, E-cadherin and p120 accumulation at CCJs was reduced, whereas a control scrambled siRNA was ineffective (Fig. 5E,F; supplementary material Movies 7 and 8). Similar results were obtained in HCT-116 and Caco-2 cells (unpublished data). All these data demonstrate that flotillins are required for N-cadherin and E-cadherin accumulation at cell–cell contacts.

To analyse the functional consequence of flotillin knockdown on cadherin-mediated cellular processes, we examined whether flotillin 1 and flotillin 2 shRNA C2C12 myoblasts could undergo myogenesis, a process that requires N-cadherin engagement (Charrasse et al., 2002). Downregulation of flotillin 1 impaired myogenesis induction, as shown by the absence of troponin T expression (Fig. 6A) and myoblast fusion (Fig. 6A,B). These effects were specific because expression of Rflot1 in flotillin 1 shRNA cells rescued troponin T expression and myoblast fusion (Fig. 6A). We next analysed the functional consequence of flotillin 1 knockdown on the MCF-7 monolayer thickness and tightness by measuring the transepithelial resistance (TER). flotillin 1 siRNA MCF-7 cells collapsed and the monolayer was less than half the thickness of parental or scrambled siRNA MCF-7 cells (Fig. 6C,D; supplementary material Movies 9 and 10). Moreover, the TER of flotillin 1 siRNA MCF-7 cells was three times lower than in scrambled siRNA MCF7 cells (Fig. 6E). Finally, we studied the effect of flotillin 1 knockdown on cell–cell adhesion by performing aggregation assays. Aggregation of C2C12 and MCF-7 cells was decreased by flotillin 1 knockdown (Fig. 6F). These data demonstrate that flotillins are required for CCJs establishment and functionality.

As flotillins are highly enriched in cholesterol-rich membrane domains (Bickel et al., 1997; Watabe-Uchida et al., 1998; Stuermer et al., 2001; Fernow et al., 2007), we assessed whether they could participate in confining cadherins in microdomains at the CCJs. First, to analyse the co-distribution of N-cadherin or E-cadherin and flotillins in cholesterol-rich membrane microdomains, we labelled GM1, a ganglioside that is associated with cholesterol-rich membrane microdomains, using fluorescent cholera toxin B (CTX-B). Cadherins and flotillins colocalised with CTX-B at CCJs in both C2C12 myoblasts and MCF-7 cells (Fig. 7A). We then analysed whether flotillins could be recruited to GM1-containing microdomains by using a technique based on the lateral cross-linking of GM1 (Ludwig et al., 2010). As expected (Stuermer et al., 2001), we confirmed that patching of GM1 with the CTX-B subunit and anti-CTX-B antibodies resulted in co-patching of flotillin 1 (Fig. 7B), thus showing that flotillin 1 is associated with GM1-containing membrane microdomains. As previously shown (Causeret et al., 2005; Taulet et al., 2009), we also observed that GM1 patching induced the co-patching of N-cadherin–GFP with GM1 in luciferase shRNA C2C12 myoblasts. In contrast, N-cadherin–GFP co-patching with GM1 was impaired in cells in which flotillin 1 was knocked down (Fig. 7C). Then, we monitored cadherin–flotillin association in conditions in which cholesterol-rich membrane microdomains were disrupted either by cholesterol chelation upon the addition of methyl-β-cyclodextrin (MCD) or by solubilisation using N-octylglucoside. Cadherin–flotillin association was impaired by addition of MCD or N-octylglucoside (Fig. 7D). In contrast, cadherin–β-catenin association remained unchanged. These data suggest that cholesterol-rich membrane microdomains are essential for the association of flotillins with N- and E-cadherin complexes.

We previously demonstrated that N-cadherin association with cholesterol-rich membrane microdomains results in its association with p120 catenin and its stabilization at CCJs (Causeret et al., 2005; Taulet et al., 2009). We thus assessed whether flotillin 1 knockdown affected p120 catenin interaction with N- and E-cadherin and found that this was the case in both mesenchymal C2C12 myoblasts and epithelial MCF-7 and HCT-116 cells (Fig. 8A; supplementary material Fig. S2B). We then analysed the role of flotillin in cadherin lateral stabilization at CCJs using fluorescent recovery after photobleaching (FRAP) experiments. We measured the diffusion coefficients (D) and the mobile fractions (M) of N-cadherin–GFP in luciferase or flotillin 1 shRNA C2C12 myoblasts (Fig. 8C; supplementary material Fig. S3). In flotillin 1 shRNA C2C12 myoblasts, the D and M values at cell–cell contacts were increased compared to those measured in luciferase shRNA C2C12 myoblasts, showing that in the absence of flotillin 1, N-cadherin–GFP was not immobilized at CCJs. Similarly, we observed an increase in D and M values for E-cadherin–GFP in flotillin 1 siRNA MCF-7 cells compared with control scrambled siRNA MCF-7 cells (Fig. 8D). These data show that flotillins are required for cadherin stabilization at CCJs.

As flotillin 1 knockdown affected p120 catenin interaction with N- and E-cadherin and previous works reported that cadherins are endocytosed in the absence of p120 catenin binding (Xiao et al., 2005; Xiao et al., 2007), we then analysed cadherin cytoplasmic distribution in flotillin 1 knockdown cells. Colocalisation of N-cadherin and E-cadherin with the LysoTraker dye, a marker of the lysosomal pathway, was increased in C2C12 myoblasts and MCF-7 cells in which flotillin 1 was knocked down in comparison with controls (Fig. 8E,F). Conversely, no increased colocalisation of cadherins with the Rab4- or Rab11-positive recycling compartment was observed (unpublished data). We then assessed the effects of flotillin 1 knockdown on N-cadherin protein expression level at the PM, and found that it was decreased in flotillin 1 shRNA C2C12 myoblasts (Fig. 8G). N- and E-cadherin expression levels were also decreased in whole cell lysates of flotillin 1 shRNA C2C12 myoblasts (Fig. 8H) as well as flotillin 1 siRNA MCF-7 (Fig. 8I) and HCT-116 cells (supplementary material Fig. S2A). Cadherin decrease could be due to degradation because, following incubation with cycloheximide to inhibit protein synthesis, N-cadherin level was lower in flotillin 1 shRNA C2C12 myoblasts than in control luciferase shRNA C2C12 cells (supplementary material Fig. S4A). However, inhibition of lysosomal activity using chloroquine, which decreases cadherin degradation, did not rescue cadherin accumulation at CCJs in flotillin 1 knockdown cells (supplementary material Fig. S4B). This suggests that the role of flotillins in CCJs formation does not involve modulation of cadherin endocytosis and degradation, in agreement with our data showing that they are required for cadherin stabilization at the PM.

These data show that cadherins and flotillins are co-distributed in cholesterol-rich membrane microdomains at CCJs and that flotillins are required for cadherin recruitment to GM1-containing membrane microdomains, which allows p120 catenin binding to cadherins and cadherin lateral stabilization at CCJs.

CCJs are highly dynamic structures and their continuous remodelling is required for the maintenance of tissue integrity, morphogenetic movements, delamination and epithelial to mesenchymal transition (Harris and Tepass, 2010). Our study shows that flotillins play a role in cadherin stabilization at cell–cell contacts and in the integrity and functionality of CCJs.

We identify flotillins as new cadherin partners in mesenchymal C2C12 myoblasts and in MCF-7, Caco-2 and HCT-116 epithelial cells. Detailed analysis of the distribution of N- and E-cadherins and flotillins in different cell lines allowed us to conclude that they are associated at the PM and in CCJs, but not in intracellular compartments. This is consistent with previous studies showing localisation of flotillins at regions of cell–cell contacts (Morrow et al., 2002; Neumann-Giesen et al., 2004; Liu et al., 2005; Glebov et al., 2006) and a role of flotillins in the regulation of CCJs during morphogenesis (Hoehne et al., 2005).

Moreover, our work highlights a new role for flotillins as major regulators of N- and E-cadherin-mediated cell–cell adhesion both in mesenchymal and epithelial cells. Functionally, we show that flotillins are required for cadherin stabilization at CCJs and formation of functional CCJs to allow cadherin-dependent signalling, such as myoblast differentiation and fusion and epithelial CCJ integrity and functionality. In agreement with our data, it was recently reported that flotillin 1 is required for E-cadherin assembly at CCJs in human colon adenocarcinoma HT-29 cells (Chartier et al., 2011). In A431 cells, flotillin 1 knockdown resulted in reduced EGFR endocytosis that induced E-cadherin macropinocytosis (Solis et al., 2012). We did not observe cadherin macropinocytosis in C2C12 myoblasts or MCF-7 and Caco-2 cells in which flotillin 1 was knocked down, suggesting that this mechanism is specific for cells that overexpress EGFR, such as the A431 cell line.

Flotillins are found in cholesterol-rich membrane microdomains and define membrane microdomains that are distinct from the caveolar PM microdomains (Bickel et al., 1997; Watabe-Uchida et al., 1998; Stuermer et al., 2001; Fernow et al., 2007). Previous studies showed that cadherin association with cholesterol-rich membrane microdomains at CCJs stabilizes cadherin (Causeret et al., 2005). Thus, flotillin–cadherin association in these membrane domains could have a role in cadherin stabilization at CCJs. Interestingly, we observed that the cadherin–flotillin complex is perturbed by treatment with N-octylglucoside and methyl-beta cyclodextrin, suggesting that this association occurs in a lipid environment enriched in cholesterol. Formation of a complex containing the PM proteins polycystins, E-cadherin and flotillin 2 in cholesterol-containing membrane domains has been reported in primary human kidney epithelial cells (Roitbak et al., 2005). Moreover, GM1 clustering by cholera toxin results in co-clustering of flotillins [(Stuermer et al., 2001), Fig. 7A] and cadherins (Causeret et al., 2005; Taulet et al., 2009; Boscher et al., 2012). We show that the cadherin–flotillin association participates in the recruitment of N-cadherin to ganglioside GM1-associated membrane microdomains. Similarly, in flotillin-1-depleted human colon adenocarcinoma HT-29 cells the recruitment of E-cadherin and p120 catenin to cholesterol-rich membrane microdomains is impaired (Chartier et al., 2011). We demonstrate that, as a consequence of the perturbation of cadherin association with GM1-containing membrane microdomains in cells in which flotillins were knocked down, p120 catenin association with cadherins, which occurs in membrane microdomains (Taulet et al., 2009), is perturbed. p120 catenin plays a major role in cadherin stabilization at cell–cell contacts (Thoreson et al., 2000; Davis et al., 2003; Xiao et al., 2003), thus explaining the inability of cadherins to be stabilized at CCJs in the absence of flotillins.

Several studies have suggested that flotillin microdomains represent assembly sites for active signalling platforms (Fecchi et al., 2006; Langhorst et al., 2006; Rajendran et al., 2008; Schneider et al., 2008; Rossy et al., 2009; Naslavsky and Caplan 2011; Zhu et al., 2012); our data argue for a role of flotillin membrane microdomains in the dynamic association of cadherins at CCJs. During CCJ establishment, cells form transient contacts, mediated by cadherins, at dynamic lamellipodia extensions. As the CCJs mature, cadherins become stabilized and concentrated in clusters between opposite cell surfaces (Harris and Tepass, 2010). Here we show that, although flotillins are constitutively associated with cadherins at the PM, their colocalisation with cadherins increases during cell–cell contact stabilization and maturation. This is in agreement with the ability of flotillin to oligomerise and to form microdomain scaffolds (Neumann-Giesen et al., 2004; Stuermer, 2010) and supports the hypothesis of a role for flotillin microdomains in cadherin clustering and stabilization at mature CCJs. A role for flotillin in the lateral assembly of transmembrane protein signalling complexes was described during neutrophil and lymphocyte T uropod formation (Gasser et al., 2006; Rossy et al., 2009; Ludwig et al., 2010; Affentranger et al., 2011) and also in EGFR and amyloid precursor protein clustering (Schneider et al., 2008; Naslavsky and Caplan, 2011). Several findings suggest that flotillins interact with and modulate the cortical actin cytoskeleton (Langhorst et al., 2007; Ludwig et al., 2010) and flotillin interaction with F-actin is also needed for the formation of the uropod in T lymphocytes (Affentranger et al., 2011). The actin cytoskeleton is a key player during the aggregation of cadherin–catenin clusters at CCJs (Yonemura et al., 1995; Watabe-Uchida et al., 1998; Vasioukhin and Fuchs, 2001; Harris and Tepass, 2010; Smutny et al., 2010; Taguchi et al., 2011; Brieher and Yap, 2013; Hong et al., 2013; Michael and Yap, 2013) and flotillin–F-actin cytoskeleton association/regulation might participate in cadherin stabilization at CCJs. Along this line, super-resolution structured illumination microscopy, because of its high resolution (approaching 100 nm), showed that cadherin-–flotillin complexes at CCJs are linked with the F-actin cytoskeleton.

In summary, our experiments reveal that flotillin microdomains participate in cadherin stabilization at cell–cell contacts and are essential for CCJs formation. Flotillins are required for cadherin recruitment to GM1-containing membrane microdomains thus allowing efficient p120 catenin recruitment, and as a consequence, stabilizing cadherins at CCJs and protecting cadherins from endocytosis and degradation (Ishiyama et al., 2010; Nanes et al., 2012).

It will now be interesting to investigate the role of flotillins during CCJs remodelling, for instance during embryonic development and morphogenesis and also cancer development.

The N-cad-GFP and N-cadAAA-YFP constructs were previously described (Taulet et al., 2009). The flotillin (Flot) Flot2-GFP, Flot2R1MCT-GFP, Flot1-mCherry, Flot1-HA and Flot2-HA constructs were from V. Niggli (University of Bern, Switzerland) (Rossy et al., 2009; Affentranger et al., 2011). Flot2-mCherry was made by cloning Flot2 in the pmCherry-N1 vector (Clontech) using the XhoI and EcoRI sites. RFlot1-mCherry and RFlot1-GFP were made by introducing two silent mutations with the QuikChange XL mutagenesis kit (Stratagene) using the following primers: 5′-GCATCCAACAGATCCAAAGGATATCTCTCAACACACTGACC-3′ and 5′-GGTCAGTGTGTTGAGAGATATCCTTTGGATCTGTTGGATGC-3′. The deletion mutant R1L1 of flotillin 2 (Flot2R1L1) was previously described (Langhorst et al., 2006) and lacks the C-terminal half of the protein (amino acids 278–428), including the EA repeat region, predicted to form coiled coils that are required for flotillin oligomerisation. The DNA sequence of the R1L1 mutant was amplified by PCR using full-length mouse flotillin 2 as a template and the following primers: 5′-ctcgagATGGGCAATTGCCACACGG-3′ and 5′-gaattcCACGCAGGATCTCCTGTGCC-3′, introducing restriction sites for XhoI and EcoRI. PCR fragments were subcloned into the pmCherry-N1 vector (Clontech). Cells were transfected with the different constructs using the JetPEI Transfection Reagent (Polyplus Transfection, Illkirch, France).

The short interfering RNA (shRNA) constructs flotillin 1 shRNA and flotillin 2 shRNA were made by inserting the oligonucleotide 5′-gatccgTCCAGAGGATCTCTCTCAAttcaagagaTTGAGAGAGATCCTCTGGACttttttacgcgtG-3′ (bold letters correspond to oligonucleotide 113–131 of the mouse Flot1 cDNA sequence, NM_008027.2), or the oligonucleotide duplex 5′-gatccgGGTGAAGATCATGACGGAGttcaagagaCTCCGTCATGATCTTCACCcttttttacgcgtg-3′ [bold letters correspond to oligonucleotides 222–240 of the mouse Flot2 cDNA sequence, NM_CAI25705; (Munderloh et al., 2009)] into the retroviral vector pSIREN-RetroQ according to the manufacturer's protocol (BD Biosciences). Luciferase shRNA was used as control (Fortier et al., 2008). Retrovirus production in Phoenix cells, infection and selection were performed as described previously (Fortier et al., 2008). Cells were grown continuously in 1 µg/ml puromycin.

For RNA-mediated interference (RNAi) experiments to knock down flotillin 1, cells were transfected using Interferin (Polyplus Transfection) with a combination of the two siRNA oligonucleotide duplexes (Eurogentec, Belgium) 5′-GUGGUUAGCUACACUCUGA-3′ (Ludwig et al., 2010) and 5′-CACACUGCCCUCAAUGUC-3′ (Glebov et al., 2006), and used in experiments 72 hours after transfection. As a control, a scrambled siRNA was used (Dharmacon, ref: D-001810-01-05).

Mouse antibodies used were against flotillin 1, flotillin 2, N-cadherin, E-cadherin, α-, β- and p120 catenin (Becton Dickinson), GFP (Roche), troponin T (Sigma-Aldrich) and transferrin receptor (Zymed). Rabbit antibodies were against caveolin (BD), flotillin 2 (Cell Signaling), p120 catenin (Santa Cruz) and actin (Sigma-Aldrich). CD44 was stained using a rat monoclonal antibody (Becton Dickinson). Alexa-Fluor-350, -488, -546, -563-conjugated antibodies and Alexa Fluor 647 cholera toxin B (CTX-B) were from Invitrogen. Horseradish-peroxidase-conjugated antibodies were from Amersham Biosciences. IRDye 680/800-conjugated antibodies were from Thermo Scientific. Coumarin-isothiocyanate-conjugated phalloidin, Hoechst (0.1 µg/ml), EGTA (2 mM), cycloheximide (10 µg/ml) and MCD (4 mM) were from Sigma-Aldrich. LysoTraker DND99 (Molecular Probes) was used at 50 nM. Chloroquine (Sigma) was used at 100 µM.

Human epithelial HCT-116 cells, MCF-7 and Caco-2 cells and mouse L cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS). Culture and differentiation of C2C12 mouse myoblasts were as previously described (Charrasse et al., 2006). To rescue myoblast differentiation in flotillin-1-shRNA-transfected cells, flotillin 1 shRNA C2C12 myoblasts were transfected with RFlot1–mCherry or RFlot1–GFP and sorted using fluorescence activated cell sorting and then selected for resistance to G418 (100 µM). Then, they were cultured in differentiation medium for 4 days. Cell aggregation assays were performed as described previously (Causeret et al., 2005).

For TER experiments, MCF-7 cells were plated in duplicate onto polycarbonate membrane (0.4 µm pore size) inserts (diameter: 6.5 mm; Transwell, Costar) and transfected with flotillin 1 or scrambled siRNA duplexes. TER measurements were performed 48 and 96 hours post-transfection using the Millicell ERS-2 system (Millipore) according to the manufacturer's protocol.

PM-enriched fractions were prepared from C2C12 or L cells as previously described (Gauthier-Rouvière et al., 1998) except that cells were lysed in hypotonic buffer with a tissue grinder without quick freezing. PM fractions were then resuspended in IP buffer (10 mM Pipes pH 7, 100 mM NaCl, 300 mM sucrose, 3 mM MgCl₂, 0.5% IGEPAL CA-640, 1 mM EDTA, 1 mM Na₃VO₄, protease inhibitor cocktail) through a 25 gauge needle. A portion of the PM fraction corresponding to 1 mg of protein was then immunoprecipitated using 1 µg of relevant or irrelevant monoclonal antibodies (4°C for 2 hours) followed by incubation with protein-G–Sepharose 4 Fast Flow (GE Healthcare; 4°C for 1 hour). Alternatively, PM fractions were treated with 60 mM N-octylglucoside (Sigma Aldrich) prior to addition of the antibodies.

Protein concentrations in whole-cell extracts and PM fractions were determined with the BCA protein assay kit (Pierce). Proteins from whole cell extracts and PM fractions (20–30 µg) and from IP samples were resolved on 10% polyacrylamide gels and transferred to Immobilon-FL membranes (Millipore). Primary and secondary antibodies were diluted in blocking buffer (Rockland) containing 0.1% Tween 20. Detection and analysis were performed using either the Odyssey Infrared imaging system (LI-COR Biosciences) or the Western Lightning Plus-ECL kit (Perkin Elmer). Immunoblots were quantified by densitometry using Odyssey V3.0 and ImageJ.

Cells were fixed in 3.2% paraformaldehyde (in phosphate-buffered saline; PBS) for 15 minutes, followed by a 2-minute permeabilisation in 0.1% Triton X-100 (in PBS) and saturation in 2% BSA (in PBS). Cells were incubated with primary and secondary antibodies in PBS containing 2% BSA. Experiments of ganglioside GM1 patching were performed on parental, luciferase or flotillin 1 shRNA C2C12 myoblasts expressing either Flot1–GFP or N-cad–GFP as described (Causeret et al., 2005).

Images were taken with either a Zeiss LSM Meta 510 or a Leica SP5 confocal microscope or with a Metamorph-driven (Molecular Devices, Sunnyvale, CA) wide-field Axioimager Z2 fluorescence microscope (Zeiss, Germany) with a PL APO 63× objective (NA 1.32, Leica, Melville, NY) and a Coolsnap HQ camera (Photometrics, Woburn, MA). Stacks of confocal images were then processed with Imaris (Bitplane, Zurich, Switzerland) for visualization and volume rendering. Colocalisations of protein fluorescence were analysed with the Imaris Colocalisation module.

Images were processed using Adobe Photoshop and assembled using Adobe Illustrator.

To label lysosomes, transfected cells were incubated with LysoTraker DND99 (Molecular Probes) and then imaged at 37°C with a Plan Fluor 40× 1.3 NA objective using an inverted TE-2000 Nikon microscope. Cells were illuminated with 491 nm and 561 nm lasers for better power stability. Fluorescence images were collected through a double band FF498/581 dichroic mirror, with a 510AF20 emission filter for GFP and a 630AF60 emission filter for Cherry. Time series were acquired with an EMCCD camera (Cascade), which was controlled by MetaMorph. Time series of captured images were deconvolved, and restored images saved as TIFF files that were compiled into QuickTime movies using MetaMorph.

Sixteen-bit images were captured and epifluorescence images were first restored with Huygens (Scientific Volume Imaging b.v., Hilversum, The Netherlands) (Ponti et al., 2007). Huygens is an iterative program that can reassign light, after encoding as grey levels, to its sources in the stack with a very high probability using a point spread function (PSF). This process removes the fuzziness contained in the stack, while keeping the 3D information. In the present study the maximum likelihood estimation (MLE) algorithm was used throughout. Restored stacks were further processed with Imaris (Bitplane, Zurich, Switzerland). Colocalisation was analysed with the Imaris Colocalisation module.

C2C12 myoblasts were cultured on high precision no. 1.5H coverslips (Marienfeld GmbH) and transfected with HA-tagged flotillin 1 or 2. Following incubation with rabbit anti-HA (Zymed) and anti-N-cadherin (Becton Dickinson) antibodies and then with Alexa-Fluor-594-conjugated anti-rabbit and Alexa-Fluor-488-conjugated anti-mouse antibodies and Alexa-Fluor-350-conjugated phalloidin (Molecular Probes), samples were mounted using Prolong Gold (Life Technologies). 3D-SIM imaging was performed using an OMX microscope (Applied Precision Inc.) equipped with 405 nm, 488 nm and 592 nm lasers and the corresponding dichroic and filter sets. Reconstruction and alignment of the 3D-SIM images was carried out with softWoRx v 5.0 (Applied Precision Inc.). 100 nm green fluorescent beads (Invitrogen) were used to measure the optical transfer function (otf) used for the 405 and 488 channels, and 100 nm red fluorescent beads (Invitrogen) were used to measure the otf used for the 592 channel. 170 nm TetraSpeck beads (Invitrogen) were employed to measure the offsets and rotation parameters used in the realignment.

FRAP recoveries were acquired at 37°C on N-cad–GFP-or E-cad–GFP-expressing cells plated on glass Petri dishes. Lateral diffusion coefficients and mobile fractions of N-cad–GFP were measured by FRAP using a Zeiss LSM Meta 510 confocal microscope as described previously (Causeret et al., 2005). Lateral diffusion coefficients (D) and the mobile fractions (M) of E-cad–GFP were measured by FRAP using the Zeiss LSM 780 confocal microscope. The 488-nm line of the argon laser was used for the excitation of GFP. Cells were observed using a 63× oil immersion objective (1.4 numerical aperture) with a pinhole adjustment resulting in a 0.9 µm optical slice (1 airy unit). After five pre-bleach scans (1 scan/0.5 seconds), a region of interest (ROI) with a radius of 1 µm was bleached, and fluorescence recovery was sampled for 1 minute, once every 0.5 seconds. Half-recovery times (t_1/2) and mobile fractions (M) were determined by fitting the normalized recovery curves using the equation given in Axelrod et al. with the Zeiss Zen Software (Axelrod et al., 1976). D was then calculated using the equation D = w2/4t_1/2. Paired Student's t-tests were used for statistical analysis.

All data are the means ± s.e.m. of 3–7 replicates. Statistical analysis was performed using the Mann–Whitney U-test and the paired Student's t-test for FRAP data and line-scan data.

We thank Verena Niggli for providing plasmids encoding flotillins; and Sylvain de Rossi, Virginie Georget and Julio Mateos Langerak at the Montpellier Imaging (http://www.mri.cnrs.fr/) and the Montpellier Proteomic Facilities (http://www.fpp.cnrs.fr/).