The focal adhesion protein PINCH-1 associates with EPLIN at integrin adhesion sites

The skin is composed of an epidermal and a dermal compartment, which are separated by a basement membrane. The primary cell types of the epidermis are keratinocytes, which form different cell layers. The basal cell layer is attached to the basement membrane, displays high proliferative capacity and harbors epidermal stem and progenitor cells. When basal keratinocytes initiate differentiation they exit the cell cycle, detach from the basement membrane and move to the suprabasal layers where they eventually die and slough off as enucleated dead cells (squames).

Adhesion of keratinocytes to the basement membrane and to each other is essential for epidermal development and homeostasis. Keratinocyte adhesion to the underlying basement membrane is mainly mediated by integrins (Watt, 2002). Genetic and cell biological studies have revealed that integrins fine tune multiple functions of basal keratinocytes including basement membrane attachment, proliferation, differentiation, migration and growth factor and cytokine signaling (Dowling et al., 1996; van der Neut et al., 1996; Brakebusch et al., 2000; Raghavan et al., 2000; Grose et al., 2002; Sehgal et al., 2006; Rodius et al., 2007; Margadant et al., 2009; Niculescu et al., 2011). The short integrin cytoplasmic domains lack enzymatic activity for signal transduction, which makes them dependent on the recruitment of adaptor and signaling proteins that assemble into large signaling hubs called focal adhesions (FAs). Integrin-linked kinase (ILK) is an adaptor protein that is recruited to integrin β tails upon integrin–ligand binding. ILK forms a heterotrimeric complex with PINCH and parvin called the IPP complex. The IPP complex is pre-assembled in the cytoplasm and enables FA targeting, stability of its partners, signaling to the actin cytoskeleton and crosstalk with growth factor signaling pathways (Legate et al., 2006; Wickström et al., 2010a; Azimifar et al., 2012). Absence of ILK in basal keratinocytes of skin leads to epidermal adhesion defects and hair loss (Lorenz et al., 2007; Nakrieko et al., 2008). The role of PINCH in skin is not known. There are two PINCH proteins (PINCH-1 and PINCH-2, also known as LIMS1 and LIMS2, respectively); each has five LIM domains followed by a C-terminal nuclear localization signal. The first LIM domain of PINCH-1 and PINCH-2 binds ILK (Zhang et al., 2002; Braun et al., 2003). PINCH-1 is ubiquitously expressed, whereas PINCH-2 is enriched in heart and skeletal muscle and absent in keratinocytes. Deletion of the PINCH-1 gene in mice leads to peri-implantation lethality associated with impaired cell–matrix and cell–cell adhesion (Li et al., 2005; Liang et al., 2005). Deletion of PINCH-2 produces no obvious phenotype (Stanchi et al., 2005). It is believed that PINCH-1 regulates cell adhesion stability and maintenance through ILK, parvin and actin regulatory proteins that associate with the LIM2–LIM5 domains (Tu et al., 1998; Tu et al., 2001; Bock-Marquette et al., 2004; Chiswell et al., 2008; Chiswell et al., 2010).

Keratinocytes develop multiple cell–cell adhesion sites, including adherens junctions, which regulate cell polarity and adhesion between two neighboring cells (Jamora and Fuchs, 2002). Adherens junctions form through homophilic interactions between E-cadherins whose cytoplasmic domains are associated with a filamentous actin belt through a large number of adaptor proteins (Abe and Takeichi, 2008). Recently, epithelial protein lost in neoplasm (EPLIN, also known as LIMA1) has been shown to serve as an adaptor in adherens junctions and to form a ternary complex with α-catenin and F-actin without impairing their binding to β-catenin (Gates and Peifer, 2005; Mège et al., 2006; Abe and Takeichi, 2008; Baum and Georgiou, 2011). EPLIN was originally identified in a screen designed to identify proteins that are significantly downregulated or lost in human cancer cells (Maul and Chang, 1999). The EPLIN protein consists of two actin-binding domains (ABDs) separated by a central LIM domain. Two alternatively spliced EPLIN mRNAs have been described, which are translated into a 90 kDa and 110 kDa protein. The function of EPLIN is poorly understood. It has been shown that the two ABDs bundle filamentous actin (Maul et al., 2003; Schaller et al., 2013) and that ERK1/2 is able to phosphorylate EPLIN, which inhibits F-actin binding resulting in the translocation of EPLIN to the cell periphery and dorsal membrane ruffles (Han et al., 2007). Moreover, EPLIN has recently been shown to localize to FAs, where it interacts with paxillin and possibly stabilizes FAs (Tsurumi et al., 2014).

To directly analyze PINCH-1-regulated cell adhesion in epidermis and hair follicles, we deleted the PINCH-1 gene in basal keratinocytes using the Cre/loxP system. The mutant mice suffer from epidermal blisters and hyperthickening, progressive hair loss and cell–matrix adhesion defects. PINCH-1-deficient keratinocytes display severe adhesion, spreading and migration defects. Immunoprecipitation of PINCH-1 from keratinocyte lysates combined with mass spectrometry identified EPLIN as a new PINCH-1 interaction partner. Cell biological studies with primary keratinocytes revealed that EPLIN recruitment to cell–matrix adhesion sites is controlled by PINCH-1, whereas PINCH-1 recruitment to FAs is not grossly affected in EPLIN-depleted cells. The implications of our findings are discussed.

To directly study the function of PINCH-1 in the epidermis we crossed mice carrying a loxP-flanked PINCH-1 gene (PINCH-1^fl/fl) (Li et al., 2005) with mice expressing the Cre recombinase under control of the keratin 5 (K5, also known as KRT5) promoter (K5-Cre) (Ramirez et al., 2004). The number of mice with the floxed PINCH-1 gene and the K5-Cre transgene was low (one out of 45 offspring) suggesting that the two genes reside in close proximity on the same chromosome. PINCH-1^fl/wt mice with and without the K5-Cre transgene were normal and served as controls (control). Mice with two floxed PINCH-1 alleles and the K5-Cre transgene (P1-K5) were viable (Fig. 1A). Western blotting (WB) of epidermal lysates and immunostaining of back skin from P1-K5 mice with antibodies that recognizes either PINCH-1 or both PINCH-1 and PINCH-2 revealed an almost complete loss of PINCH protein, which was accompanied by diminished ILK levels (supplementary material Fig. S1A–D). Although P1-K5 mice were normal at birth, their hair appeared shaggy with small areas of alopecia appearing at postnatal day 14 (P14) (Fig. 1A). By P56 P1-K5 mice had lost their hair and developed a patchy pigmentation of their skin (Fig. 1A). Given that PINCH-2 was not expressed de novo in P1-K5 keratinocytes (supplementary material Fig. S1B,C), the low levels of PINCH-1 protein in P56 epidermal lysates indicates that cells escaping K5-Cre-mediated PINCH-1 gene deletion expanded in the epidermis of P1-K5 mice. To avoid the presence of PINCH-1-expressing cells in our analyses, all skin histology and cell biology studies with primary keratinocytes were conducted with P1-K5 mice that were 2 weeks of age or younger.

Hematoxylin-eosin (H&E) staining of the back skin revealed that the P14 skin of P1-K5 mice contained sparse and abnormal hair follicles, a hyperthickened interfollicular epidermis (IFE) and small blisters at the dermal–epidermal junctions (DEJ) (Fig. 1B,C). When quantified, the numbers of blisters per millimeter skin were significantly (P<0.001) higher in P1-K5 skin (supplementary material Fig. S1E). The proliferating cells determined by Ki67 staining were 50% of the total basal cell number both in control and P1-K5 epidermis. Interestingly, 20% of the total cell number in the suprabasal layers of P1-K5 epidermis was also Ki67+, which was observed in less than 5% of the suprabasal cells in the control skin (supplementary material Fig. S1F), pointing to an adhesion defect and loss of proliferating P1-K5 basal keratinocytes. Hair follicle morphogenesis in control mice was complete at P14 and all hair follicles resided deep in the subcutis (Fig. 1B, left panel). In contrast, P1-K5 skin exhibited two types of hair follicles; fully developed hair follicles with hyperthickened outer root sheaths (ORS) reaching the subcutis, and short hair follicles with an abnormal shape residing in the dermis and with either a distorted or no dermal papilla attached to them (Fig. 1B, right panel; supplementary material Fig. S1G, right panel).

The deposits of laminin 332 (Lm322) and nidogen, two major components of the epidermal–dermal basement membrane, and the expression of α6 integrin appeared as continuous layers between epidermis and dermis of control skin. In P1-K5 skin the Lm322 and nidogen deposits and α6 integrin localization were also frequently detected on the apical and lateral sides of basal keratinocytes (Fig. 1D,E; supplementary material Fig. S1H). F-actin, the cell–cell adhesion protein E-cadherin, as well as desmosomal plaque proteins desmoplakin and plakoglobin localized to the subcortical lateral and apical regions of basal keratinocytes in control skin. In basal P1-K5 keratinocytes, however, E-cadherin and F-actin extended to the basal sides suggesting that cell polarity is distorted (Fig. 1D,E). These findings demonstrate that PINCH-1 plays a role in cell–matrix adhesion stability, basement membrane integrity and keratinocyte polarity.

PINCH and ILK are obligate binding partners (Legate et al., 2006), and loss of ILK severely compromises integrin-mediated functions of keratinocytes (Lorenz et al., 2007). To compare cell–matrix adhesion and spreading between P1-K5 and ILK-deficient keratinocytes (ILK-K5), we derived primary keratinocytes from both mouse strains and performed plate and wash assays by seeding them on laminin-322 (Lm322), fibronectin or collagen I (Col1). As expected, control keratinocytes readily adhered to the three matrix substrates, whereas adhesion of P1-K5 as well as ILK-K5 keratinocytes to Lm332, fibronectin and Col1 was significantly (P<0.001) impaired (Fig. 2A). The P1-K5 and ILK-K5 keratinocytes that were able to adhere showed severely impaired spreading, even 20 h after seeding on a mixture of fibronectin and Col1 matrix (Fig. 2B–D). Interestingly, the spreading defect of P1-K5 cells was more pronounced when compared to ILK-K5 cells (Fig. 2B). Immunostaining of control cells revealed numerous FAs containing ILK and PINCH that were linked to F-actin stress fibers (Fig. 2C,D). ILK-K5 cells developed paxillin-positive FAs linked to thin actin fibers and thick subcortical actin deposits (Fig. 2C). P1-K5 cells also formed small, paxillin-positive adhesion sites and accumulated a thick layer of F-actin beneath the cell cortex (Fig. 2D). Given that cell adhesion and spreading are important parameters of cell migration, we analyzed migration of P1-K5 keratinocytes by performing scratch wound assays and found that control keratinocytes closed the wound within 12–24 h, whereas P1-K5 cells were unable to move into the wounded area (Fig. 2E). To exclude that the decreased P1-K5 keratinocyte migration, spreading and adhesion result from altered integrin expression and/or activation, we isolated primary keratinocytes, determined integrin levels and the level of the 9EG7 epitope, which binds the active conformation of β1 integrin (Lenter et al., 1993). The experiments revealed that the levels of all integrins tested as well as the levels for the 9EG7 epitope were similar between control and P1-K5 cells (supplementary material Fig. S2A). Taken together, these results demonstrate a strong role for PINCH-1 in cell adhesion, spreading and migration.

Although E-cadherin and F-actin distribution were altered in basal keratinocytes in vivo, cell–cell adhesion formation induced with Ca²⁺ occurred in primary (P1-K5) as well as immortalized PINCH-1^−/− keratinocytes when they were plated on a mixed fibronectin and Col1 matrix (Fig. 2F). The F-actin distribution resembled the in vivo situation with large deposits at the basal side (Fig. 2F; supplementary material Fig. S2B) and the z-axis views revealed an increased cell thickness, irregular cell shapes and stacking (Fig. 2F, lower strips). These findings and the absence of skin barrier defects (data not shown) indicate that loss of PINCH-1 expression in keratinocytes enables the formation of cell–cell adhesion sites but severely impairs the formation and functions of cell–matrix adhesion sites.

To dissect mechanistically how PINCH-1 supports keratinocyte adhesion, spreading and migration, we screened for new interaction partners by performing label-free pulldown experiments with immortalized PINCH-1^−/− keratinocytes rescued with an EGFP-tagged PINCH-1 (PINCH1^rescued) followed by mass-spectrometry (supplementary material Fig. S2C). The PINCH-1^−/− cells were derived from immortalized PINCH-1 fl/fl keratinocytes and subjected to adenoviral Cre transduction. They phenocopied the severe defects of primary P1-K5 cells, and reconstitution with EGFP-tagged PINCH-1 restored adhesion, spreading, F-actin distribution and expression of ILK and parvin (supplementary material Fig. S2D–F).

Among the proteins that were precipitated with PINCH-1 we identified known PINCH-1 interactors such as ILK and Rsu1 as well as new binding partners such as EPLIN (Fig. 3A; supplementary material Table S1). Given that EPLIN was among the top hits in our interactome screen and has also been identified as a mechanosensitive component of focal adhesions in recent proteomic screens (Schiller et al., 2011; Schiller et al., 2013), we further characterized the role of the PINCH-1–EPLIN interaction. Consistent with the proteomics data, endogenous EPLIN was co-immunoprecipitated with GFP–PINCH-1 (Fig. 3B), and endogenous PINCH-1 co-immunoprecipitated with EPLIN–GFP (Fig. 3C). Importantly, immunoprecipitation of endogenous PINCH-1 also co-precipitated endogenous EPLIN (Fig. 3D), clearly demonstrating that the two endogenous proteins associate in keratinocytes. Interestingly, α-catenin, a well-known EPLIN-binding partner (Abe and Takeichi, 2008), co-precipitated with EPLIN–GFP expressed in control or PINCH-1^−/− keratinocyctes indicating that EPLIN binding to α-catenin is independent of PINCH-1. Finally, we performed proximity ligation assays (PLAs) in PINCH-1^fl/fl and PINCH1^rescued keratinocytes. We observed no PLA signal when the assay was performed with no antibody or only with anti-EPLIN or anti-PINCH-1 antibodies (supplementary material Fig. S3A). However, when cells were incubated with both anti-EPLIN and anti-PINCH-1 antibodies the PLA signal was strong and colocalized with paxillin (Fig. 3E; supplementary material Fig. S3A) indicating that EPLIN and PINCH-1 reside in close proximity in FAs.

In agreement with previous reports (Abe and Takeichi, 2008), we also found EPLIN together with E-cadherin at cell–cell adhesion sites of densely seeded and Ca²⁺-treated control keratinocytes (Fig. 4A). In Ca²⁺-treated P1-K5 cells EPLIN was found at cell–cell adhesion sites and dispersed through the cytoplasm in vitro (Fig. 4A) and in vivo (Fig. 4B; supplementary material Fig. S3B). When untreated control keratinocytes were sparsely seeded on a fibronectin and Col1 matrix we observed a strong colocalization of EPLIN with paxillin in FAs (Fig. 4C). In contrast, EPLIN was absent from FAs of sparsely seeded PINCH-1^−/− keratinocytes and instead accumulated in the cytoplasm (Fig. 4C). Similarly, skin sections of P1-K5 mice showed regions of poor EPLIN localization at the basal side of basal keratinocytes and abnormal accumulations in the cytoplasm, whereas EPLIN at sites of cell–cell interaction was not grossly altered (Fig. 4B; supplementary material Fig. S3B). Interestingly, western blotting revealed slightly reduced EPLIN protein levels in P1-K5 keratinocytes (Fig. 4D) suggesting that the PINCH-1–EPLIN interaction stabilizes the EPLIN protein. To determine whether the reduced EPLIN levels caused the reduced EPLIN staining in FAs and the impaired spreading, we increased EPLIN levels by overexpressing EGFP-tagged EPLIN in PINCH-1^−/− keratinocytes and found that overexpressed EPLIN neither localized to cell–matrix adhesions nor rescued the spreading defect (Fig. 4E).

Next, we investigated whether PINCH-1 recruitment to integrin adhesion sites was dependent on EPLIN. To this end, we efficiently depleted EPLIN in control cells using specific small interfering RNAs (siRNAs) and found that ILK and PINCH-1 levels became slightly reduced (Fig. 5A) and E-cadherin levels decreased at AJs (supplementary material Fig. S3C). Despite the presence of substantial levels of PINCH-1 and ILK in FAs of EPLIN-depleted cells, their spreading and migration in wound closure assays were impaired (Fig. 5B–D). Taken together, these findings indicate that EPLIN is a new PINCH-1 interacting protein that is recruited to FAs in a PINCH-1-dependent manner and regulates adhesion, spreading and migration.

To investigate whether the recruitment of EPLIN to FAs requires an intact IPP complex, we isolated primary ILK-K5 keratinocytes and immunostained them for EPLIN (Lorenz et al., 2007). The poorly spreading ILK-K5 keratinocytes contained large FAs that were positive for EPLIN (Fig. 6A) suggesting that the remaining PINCH-1 protein in ILK-deficient cells is sufficient to support EPLIN recruitment to FAs independently of an intact IPP complex (Lorenz et al., 2007). Although western blotting showed low levels of PINCH in ILK-K5 cells (Fig. 6B), immunostaining failed to detect these low PINCH-1 levels in FAs owing either to the decreased protein levels and/or to the quality of the anti-PINCH antibodies (Fig. 6C). To circumvent immunostaining with antibodies, we expressed GFP-tagged PINCH-1 and Cherry-tagged paxillin cDNAs in control and ILK-K5 cells and performed total internal reflection fluorescence (TIRF) microscopy to visualize EGFP–PINCH-1 in FAs. The experiments revealed that GFP-tagged PINCH-1 indeed colocalized with paxillin in FAs of ILK-K5 keratinocytes seeded on a fibronectin and Col1 matrix (Fig. 6D). Taken together, these data indicate that PINCH-1 can localize to ILK-deficient FAs and that the low PINCH-1 levels in ILK-K5 keratinocytes are sufficient to recruit EPLIN to these sites.

In this study, we deleted the PINCH-1 gene in keratinocytes of mice and showed that loss of PINCH-1 expression leads to blisters at the sub-epidermal basement membrane, epidermal hyperplasia and impaired downward growth of hair follicles. Similar defects develop when the β1 integrin or ILK genes are deleted in keratinocytes of mice, and thus this finding supports the common view that ILK and PINCH-1 are essential adaptors downstream of integrins. A careful comparison of keratinocytes from mice lacking either ILK or PINCH-1 revealed that loss of PINCH-1 induces more severe adhesion, spreading and migration defects than loss of ILK.

PINCH-1 is believed to perform its functions in association with ILK and parvin. Contrary to the view that members of the IPP complex require each other for stability, deletion of either IPP component does not result in a complete destabilization of the remaining binding partners, as their low levels can be detected by western blotting or mass spectrometry (Lorenz et al., 2007; Fielding et al., 2008a; Fielding et al., 2008b). It is likely that IPP proteins not incorporated into the IPP complex have IPP-independent functions, either within or outside integrin adhesion sites. Indeed, ILK for example, has been shown to localize to centrosomes without PINCH or parvin where it controls spindle organization in mitosis (Fielding et al., 2008a; Fielding et al., 2008b). There is no report of PINCH-1 so far, that demonstrates a subcellular localization of a PINCH-1 fraction that is not destabilized upon loss of ILK deletion. However, it is clear from our study that low levels of PINCH-1 remain in ILK-deficient keratinocytes, although they are difficult to visualize with the available antibodies. To better visualize these low PINCH-1 levels, we tagged PINCH-1 with EGFP, expressed the fusion protein in ILK-deficient cells and observed it in integrin adhesion sites using TIRF microscopy. Although the majority of the PINCH-1 pool was recruited to FAs in an ILK-dependent manner, a small pool of PINCH-1 could be found in FAs in the absence of ILK. The ILK-independent role of PINCH-1 at FAs might explain why the spreading and adhesion defects of PINCH-1-null cells are more severe than those of ILK-null keratinocytes. How the ILK-independent pool of PINCH-1 is recruited to FAs is not clear.

How is the ILK-independent PINCH-1 operating at FAs? To address this question, we performed pulldown experiments followed by mass spectrometry and identified EPLIN as a new PINCH-1-interacting protein. EPLIN is known to bind to α-catenin in cell–cell junctions (Abe and Takeichi, 2008) and is present in the adhesome of fibroblasts (Schiller et al., 2011; Schiller et al., 2013) and FAs of mesangial cells (Tsurumi et al., 2014). Whether the interaction between PINCH-1 and EPLIN is direct or indirect is not clear, as our attempts to purify recombinant PINCH-1 for in vitro binding assays were unsuccessful. However, we made several observations pointing to a functional interdependence of PINCH-1 and EPLIN in FAs. First, proximity ligation assays revealed that PINCH-1 and EPLIN reside in close proximity in FAs. Second, PINCH-1 promotes EPLIN stability and induces the recruitment of EPLIN to FAs. Western blotting and immunostaining demonstrated that the levels of EPLIN in whole-cell lysates and the levels of EPLIN in FAs were diminished in PINCH-1-deficient keratinocytes. Although re-expression of PINCH-1 restored total EPLIN levels and EPLIN recruitment to FAs in PINCH-1-deficient keratinocytes, overexpression of EPLIN in PINCH-1-deficient cells was not sufficient to target EPLIN to FAs, indicating that EPLIN stability and FA recruitment depends on PINCH-1. Third, loss of PINCH-1 or depletion of EPLIN in keratinocytes share similar integrin-dependent defects including impaired cell adhesion, spreading and migration. Finally, our proteomics study revealed a further interesting commonality with respect to the mechanosensitivity of EPLIN and PINCH-1 (Schiller et al., 2011; Schiller et al., 2013); both are among the LIM-domain-containing proteins that require myosin-II-mediated force for their recruitment to and maintenance in FAs (Schiller et al., 2011). Interestingly, a previous paper also reported a force-dependent localization of EPLIN to cell–cell adhesion sites (Taguchi et al., 2011), where EPLIN links the E-cadherin–catenin complex to F-actin (Maul and Chang, 1999; Maul et al., 2003; Abe and Takeichi, 2008).

Given that neither loss of PINCH-1 nor depletion of EPLIN impairs the surface expression or activation state of β1 integrins, we propose that the defects caused by loss of PINCH-1 or EPLIN are due to abnormal adhesion signaling and actin dynamics. PINCH-1 might regulate adhesions and actin dynamics through ILK and/or Nck-2, which can interact with p21-activated kinase (Bokoch et al., 1996), Wiskott–Aldrich syndrome protein (Braverman and Quilliam, 1999) and DOCK180 (Tu et al., 2001), respectively. EPLIN might execute its functions through interaction partners in FAs as well as the ability to crosslink actin filaments into bundles and to fine tune Arp2/3-mediated branching of actin (Han et al., 2007; Schaller et al., 2013). EPLIN interaction partners in FAs are PINCH-1 (shown in this paper) and paxillin (Tsurumi et al., 2014). Interestingly, Tsurumi et al. (Tsurumi et al., 2014) showed that depletion of EPLIN in mesangial cells enhanced rather than reduced their migration. This is in stark contrast to our findings, which revealed a pronounced migration defect upon EPLIN depletion. An explanation for this discrepancy could be that FA signaling and/or actin remodeling are differently controlled by EPLIN and paxillin in different cell types. Keratinocytes are highly motile both in cell culture and during wound closure, whereas mesangial cells are stationary, smooth-muscle-like cells regulating intra-glomerular capillary flow and glomerular ultrafiltration (Schlöndorff and Banas, 2009). Clearly, genetic studies of EPLIN in mice and cells will cast light on its role in different cell types during development and postnatal homeostasis.

To obtain a keratinocyte-restricted deletion of the PINCH-1 gene, PINCH-1^fl/wt males expressing Cre under the control of the keratin 5 promoter were crossed with PINCH-1^fl/fl females (Li et al., 2005). Offspring were genotyped as described previously (Li et al., 2005). Primary keratinoyctes were isolated and cultured from PINCH-1^fl/fl and PINCH-1-K5 animals. All cells were immortalized by culturing primary cells long enough to let the spontaneous differentiation occur. ILK^fl/fl and ILK-K5 cells were also isolated from the respective mouse line. The PINCH-1^rescued cell line was obtained by adenoviral infection of the PINCH-1–EGFP construct, followed by cell cloning. The colonies were tested for PINCH-1–EGFP expression and the colony with an endogenous expression level was chosen for further experiments (supplementary material Fig. S2D). All animal experiments were performed according to approved guidelines of the Max Planck Society and the government of Upper Bavaria.

Primary keratinocytes were isolated and cultured as described previously (Lorenz et al., 2007). To induce differentiation and the formation of cell–cell contacts keratinocytes were grown to confluence and treated with CaCl₂ at a final concentration of 1.2 mM. Transfections were carried out with Lipofectamine 2000 transfection reagent (Invitrogen). The following expression constructs were used: PINCH-1–EGFP (Braun et al., 2003; Stanchi et al., 2005), ILK–EGFP (Sakai et al., 2003), and EGFP or Cherry-tagged EPLIN, which were cloned by inserting the mouse EPLIN cDNA into the pEGFP-N1 (Clontech Laboratories Inc., Mountain View, CA) and pCherry-N1 (kindly obtained from Roger Y. Tsien, University California San Diego, La Jolla, CA) backbone, respectively.

Adhesion of epidermal keratinocytes to extracellular matrix proteins [Poly-L-lysine (Sigma), Col1 (Cohesion), fibronectin (Merck), LM332 (provided by Monique Aumailley, University of Cologne, Cologne, Germany)] was measured as previously described (Fässler et al., 1995).

For spreading assays, cells were seeded on ColI- and fibronectin-coated dishes and allowed to spread for the indicated times. Four images were taken every 15 min during the live-cell recording. All live-cell recordings were performed at 37°C and under 5% CO₂ using a Zeiss Axiovert microscope equipped with a 10× NA0.3, 20× NA0.4, 40× NA0.6 and 100× NA1.3 objectives, a motorized scanning table (Märzhäuser) and a stage incubator (EMBL Precision Engineering, Heidelberg, Germany). Images were captured with a cooled CCD camera (Roper Scientific MicroMAX, Martinsried, Germany) using the Metamorph software (Molecular Devices, Sunnyvale, CA, USA) for microscope control and data acquisition.

TIRF images were captured with an Axiovert 200M inverted microscope (Zeiss, Oberkochen, Germany) with a 100× oil objective and a CoolSnap HQ CCD camera (Photometrics, AZ). Acquisition was controlled by Metamorph Software (Molecular Devices, Sunnyvale, CA). Images were collected at 37°C.

Immunohistochemistry of skin cryosections was carried out as described previously (Brakebusch et al., 2000). Immunostaining of cells were performed as described previously (Wickström et al., 2010b). The following antibodies and antibody dilutions were used: rat monoclonal antibody (mAb) against E-cadherin (1∶400; Zymed); rabbit polyclonal antibody (pAb) against desmoplakin (1∶500; Research Diagnostics, NV); rabbit pAb against EPLIN (1∶150; Bethyl Laboratories, Inc.); mouse mAb against ILK (1∶250; Millipore, MA); FITC-conjugated mAb against integrin α6 (1∶400; BD Pharmingen, CA, USA), rat mAb against Ki67 (1∶500; Dianova); rabbit pAb against LM332 (1∶500; obtained from M. Aumailley); rat mAb against nidogen (1∶500; Chemicon); rabbit pAb against Paxillin (1∶50; Santa Cruz Biotechnology, USA); mouse mAb against PINCH (1∶250; BD Biosciences); home-made rabbit pAb against PINCH-1 (Li et al., 2005); home-made rabbit pAb against PINCH-2; rabbit pAb against plakoglobin (1∶500; Santa Cruz, USA); rat mAb against tubulin (1∶300; clone YL 1/2, Millipore, MA, USA); mouse mAb against vinculin (1∶300; Sigma-Aldrich, MO); phalloidin–Alexa-Fluor-488 (Life Technologies, NY); Phalloidin–TRITC (Sigma-Aldrich); goat anti mouse-IgG conjugated to Cy3; goat anti rat-IgG conjugated to Cy3; goat anti rabbit-IgG conjugated to Cy3; donkey anti-rabbit-IgG conjugated to Cy3 (Jackson Immunoresearch, PA, USA); goat anti-rabbit-IgG conjugated to Alexa Fluor 488 (Sigma-Aldrich, MO); goat anti-rat-IgG conjugated to Alexa Fluor 488 (Life Technologies, NY, USA); goat anti-rat-IgG conjugated to horseradish peroxidase (HRP), goat anti-mouse-IgG conjugated to HRP and goat anti-rabbit-IgG conjugated to HRP (BioRad, CA, USA). Fluorescence images were collected by confocal microscopy (DMIRE2; Leica, Wetzlar, Germany) using the Leica Confocal Software (version 2.5 Build 1227).

The PLA kit was purchased from Olink Bioscience (Uppsala, Sweeden). Every experiment was controlled with ‘no antibody’ and ‘only one antibody’ conditions. As a representative control experiment, a ‘no antibody’ control is presented in Fig. 3E. Cells were cultured as described in the ‘Cell culture and transfection’ section. Cells were fixed with 4% PFA in PBS and stained according to the manufacturer's instructions.

Cell wounding assays were performed with monolayers of keratinocytes treated with 4 µg/ml Mitomycin C (Sigma) for 4 h and then scratched with a 200 µl plastic micropipette to obtain wound widths of 500–600 µm. Wound closure was monitored with a Zeiss Axiovert microscope equipped with 10× NA0.3, 20× NA0.4, 40× NA0.6 and 100× NA1.3 objectives, a motorized scanning table (Märzhäuser) and a stage incubator (EMBL Precision Engineering, Heidelberg, Germany) for 12 h. Images were collected with a frame rate of one every 15 min using a cooled CCD camera (Roper Scientific MicroMAX, Martinsried, Germany) and Metamorph software (Molecular Devices, Sunnyvale, CA) was used for microscope control and data acquisition. All recordings were performed at 37°C and under 5% CO₂.

Flow cytometry was carried out as previously described (Brakebusch et al., 2000). Antibodies used for FACS analysis were: FITC-conjugated hamster mAb against integrin β1; rat mAb against integrin β1 9EG7; FITC-conjugated rat mAb against integrin α6; biotinylated rat mAb against integrin αV; rat mAb against integrin β4; FITC-conjugated hamster mAb against integrin α2; biotinylated rat mAb against integrin α5 (all BD Biosciences, CA, USA); Streptavidin–Cy5 (BD Biosciences, CA, USA); mouse mAb anti-rat-IgG conjugated to FITC (BD Biosciences, CA, USA); goat anti mouse-FITC (Jackson Immunoresearch, PA, USA).

Cells were lysed in 50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1% Triton X-100, 0.05% sodium deoxycholate and 10 mM EDTA supplemented with protease inhibitors (Roche, Penzberg, Germany) and phosphatase inhibitors (Sigma), homogenized in Laemmli buffer and boiled for 5 min. Cell lysates were electrophoretically separated on SDS-PAGE gels and then transferred onto nitrocellulose membranes, followed by incubation by antibodies. Bound antibodies were detected using enhanced chemiluminescence (Millipore Corporation, Billerica).

siRNA duplexes for EPLIN and scrambled control duplexes were purchased from Sigma. Two siRNA duplexes (5′-GAAAUAGGUGCGGGUCAUU-3′ and 5′-UGUAAGCCUCACUUCAAU-3′) were transfected with Lipofectamine (Life Technologies, NY, USA) into control or PINCH-1^rescued cells. Experiments were carried out 48 h after transfection.

Cell lysis and immunoprecipitation were performed as described previously (Böttcher et al., 2012). Briefly, cells were lysed with a buffer containing 150 mM NaCl, 50 mM Tris-HCl pH 7.5, 5% glycerol, 1% IGEPAL-CA-630, 1 mM MgCl₂ and protease inhibitors. Lysates were incubated with 50 µl magnetic beads coupled with monoclonal mouse anti-GFP antibody (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany). After washes, proteins were pre-digested on the same column with trypsin, eluted from the column with DTT and further digested with trypsin overnight. Peptides were purified on C₁₈ StageTips before mass spectrometry analysis.

Peptides were separated by reverse-phase chromatography coupled to liquid chromatography tandem mass spectrometry (LC-MS/MS) analysis using an LTQ-Orbitrap Velos mass spectrometer (Thermo Fisher Scientific, MA). Peptide separation was performed with an EASY-nanoLC system (Thermo Fisher Scientific, MA) with a 180-min gradient from 5% to 35% buffer B (80% acetonitrile, 0.5% acetic acid). In the mass spectrometry method, full scans were acquired in the Orbitrap with a resolution of 60,000. The top ten most intense ions were fragmented by collision-induced dissociation (CID), and MS/MS spectra were acquired in the LTQ.

The raw files from the mass spectrometry were processed with the MaxQuant software version 1.0.13.13 (Cox and Mann, 2008). Data were searched against the forward and decoy database (pi.MOUSE.v3.62.decoy.fasta) using the Mascot search engine. Fixed modifications were carbamidomethylated cysteins and variable modifications were oxidation of methionine, and N-terminal acetylation. Maximum false discovery rates (FDR) were set to 0.01 both on peptide and protein levels. Proteins were quantified using the label-free algorithm in MaxQuant. To extract significant binders we performed a two-sample Student's t-test between triplicates, with 0.1 FDR.

Results are shown as mean±s.d. where indicated. Significance was determined by unpaired Student's t-test and where necessary a Mann–Whitney U test was also applied.