Tetraspanin CD151 and integrin α3β1 contribute to the stabilization of integrin α6β4-containing cell-matrix adhesions

The skin is our largest organ and its barrier function protects us from dehydration and pathogenic infections. To exert these functions, the skin needs to maintain its integrity, for which stable adhesion of basal keratinocytes to the underlying extracellular basal lamina is of paramount importance. This adhesion is largely mediated by hemidesmosomes (HDs), multi-protein complexes that are associated with the keratin intermediate filament system within the cell.

The integrin α6β4, a heterodimeric transmembrane protein, is a main component of HDs. In type I HDs, present in the epidermis, integrin α6β4 binds extracellularly to laminin-332 and intracellularly via two associated proteins, plectin and BP230 (BPAG1e, encoded by DST), to the keratin cytoskeleton. Besides integrin α6β4, type I HDs contain two other transmembrane proteins: the tetraspanin CD151 and BP180 (BPAG2 or collagen XVII, encoded by COL17A1) (Litjens et al., 2006; Sterk et al., 2000; Walko et al., 2015). In more simple epithelia of the intestine or the mammary gland, a rudimentary form of HDs is present. These type II HDs contain integrin α6β4 and plectin, but lack BP180 and BP230 (Fontao et al., 1997; Uematsu et al., 1994).

In humans, the importance of type I HDs for maintaining skin integrity is underscored by genetic evidence that demonstrates that mutations in any of the six genes encoding HD proteins cause the congenital inherited skin blistering disorder epidermolysis bullosa (EB) (Has and Fischer, 2018). Patients carrying null mutations in the genes for the integrin α6 or β4 subunits, or the cytolinker protein plectin, suffer from a severe form of EB, which is indicated by the absence of HDs (Ashton et al., 2001; McLean et al., 1996; Nakamura et al., 2005; Niessen et al., 1996; Vidal et al., 1995). However, HDs with grossly normal HD structures, but reduced keratin filament association, are found in patients carrying mutations in the genes for BP180 or BP230, and accordingly, these patients only suffer from a mild form of EB (Groves et al., 2010; Jonkman et al., 1995; Liu et al., 2012; McMillan et al., 1998). Mutations in CD151 have not been associated with abnormalities in HDs but nevertheless cause a mild regional skin blistering phenotype, similar to that observed in patients with mutations in the gene encoding integrin α3 (Crew et al., 2004; Has et al., 2012; Kagan et al., 1988; Vahidnezhad et al., 2017). Targeted deletion of genes encoding HDs and associated proteins in mice largely recapitulates the pathology observed in humans (Andrä et al., 1997; DiPersio et al., 1997; Dowling et al., 1996; Guo et al., 1995; Sachs et al., 2006; van der Neut et al., 1996; Wright et al., 2004).

Integrin α3β1, another laminin-binding integrin expressed by keratinocytes, is a component of focal adhesions (FAs), but can also be found together with integrin α6β4 in tetraspanin-enriched microdomains (TEMs) (also named tetraspanin webs). TEMs are formed by lateral associations of tetraspanins with one another and with other proteins. Incorporation of integrins α6β4 and α3β1 in TEMs is facilitated through binding of their α subunits to CD151 (Hemler, 2005). The association of CD151 with laminin-binding integrins has been reported to affect integrin-dependent adhesion, migration and signaling (Berditchevski, 2001; Stipp, 2010; Termini and Gillette, 2017).

While it has been shown that CD151 is colocalized with integrin α6β4 in type I HDs, the exact function of CD151 in the formation of these adhesions or other cell-matrix adhesions in keratinocytes is not known (Sterk et al., 2002). In this study, we assessed the contribution of CD151 to the formation and maintenance of laminin-associated adhesion complexes in keratinocytes. In agreement with data obtained from CD151 knockout mice and patients with CD151 mutations, our findings confirm that CD151 is not critical for the formation of type I HDs. However, we show that CD151 is essential for the maintenance of adhesion complexes in which CD151 is present together with integrin α3β1 and the HD proteins integrin α6β4 and plectin, but which are not associated with the keratin cytoskeleton.

Previous studies have shown that CD151 forms stable complexes with the laminin-binding integrins α3β1 and α6β4 (Hemler, 2005; Sterk et al., 2002). We confirmed these findings by reciprocal co-immunoprecipitation (co-IP) assays using two different monoclonal antibodies against CD151 (5C11 and 11G5) and antibodies against the integrin α3 and α6 subunits (Fig. 1A). Furthermore, these assays showed that in PA-JEB/β4 keratinocytes (immortalized keratinocytes isolated from a patient with pyloric atresia associated with junctional epidermolysis bullosa, and stably expressing wild-type integrin β4) the α6 subunit does not form a heterodimer with β1, which is consistent with the fact that integrins α3β1 and α6β4, but not α6β1, are the major laminin-binding integrins in skin (Sonnenberg et al., 1991) (Fig. 1A). Using immunofluorescence (IF) microscopy, we detected CD151 in a pattern of consecutive arches, in which it is partially co-localized with integrin α3β1. Both integrin α3β1 and CD151 are also found at the cell periphery where they do not colocalize and are present in distinctive adhesion structures (Fig. 1B). Furthermore, both integrin α3β1 and CD151 co-localized with the HD proteins integrin α6β4 and plectin (Fig. 1C,D). In summary, these results confirm that in keratinocytes CD151 binds both integrins α3β1 and α6β4, and is present in adhesion structures that contain characteristics of HDs.

To determine to what extent CD151 impacts integrin α3β1- and α6β4-containing adhesion complexes, we abrogated CD151 protein expression in PA-JEB/β4 and HaCaT keratinocytes using CRISPR-Cas9 (KO) and subsequently restored the expression of CD151 by retroviral-mediated gene transfer. Analysis by western blotting (WB) and fluorescence-activated cell sorting (FACS) confirmed the absence and restoration of CD151 expression, and showed that the deletion of CD151 did not affect the expression of integrins α3β1 and α6β4 at the plasma membrane (Fig. 2A,B). However, quantitative IF staining and confocal microscopy showed reduced clustering of integrin α3β1 (and to a lesser extent of integrin α6β4) at the basal cell surface (Fig. 2C). Furthermore, integrin α3β1 was no longer co-distributed with integrin α6β4 but showed increased co-localization with phospho-paxillin in FAs (Fig. 2D,E). In the CD151-proficient cells, integrin α6β4 was clustered in central and peripheral adhesions, but strikingly in the CD151-deficient cells it was almost exclusively present at the cell periphery (Fig. 2F–H). These results show that, while not essential for integrin α3β1 or α6β4 protein expression or their transport to the plasma membrane, CD151 is necessary for proper organization of these integrins in the central region of the cells.

To gain real-time insight into how CD151 contributes to the formation of integrin α6β4-containing adhesions in the central region of the cells, we deleted CD151 in PA-JEB keratinocytes expressing GFP-tagged integrin subunit β4 (Geuijen and Sonnenberg, 2002). The absence of CD151 and presence of β4–GFP was confirmed by FACS and WB analysis (Fig. 3A,B). The clustering of β4–GFP was followed over time in these cells by time-lapse imaging. During initial stages of cell spreading, both CD151-deficient and -proficient PA-JEB/β4–GFP cells assembled integrin α6β4-containing adhesions at the periphery of the keratinocyte islands. However, while these initial adhesion structures were maintained during later phases of cell spreading and island formation in the CD151-proficient cells, in the CD151-deficient cells, the central adhesion structures resolved and disappeared when the cells formed new peripheral adhesions after cell division and cell spreading (Fig. 3C; Movies 1 and 2). Furthermore, in fluorescence recovery after photobleaching (FRAP) experiments, a spot in the central and peripheral integrin α6β4-containing adhesions was simultaneously bleached and the recovery of the β4–GFP signal at these spots was followed over time (Fig. 3D,E). The results show that the mobile fraction of integrin α6β4 in the central integrin α6β4 adhesions is larger than in the peripheral ones, indicating a higher dynamic nature of integrin α6β4 in the central adhesions (Fig. 3E,F).

We reasoned that the observed differences in the dynamics of integrin α6β4 between central and peripheral adhesion clusters could be due to the association of integrin α6β4 with different proteins. Analysis of the distribution of laminin-332 and plectin by total internal reflection fluorescence (TIRF) microscopy showed that both proteins are associated with central and peripheral integrin α6β4-containing adhesions, with laminin-332 being slightly more abundant under the central adhesion clusters and plectin at the peripheral adhesion clusters (Fig. 4A,B). Strikingly, keratin and BP230 (which, like plectin, can connect integrin α6β4 to the keratin cytoskeleton) were associated solely with integrin α6β4 in the peripheral adhesions (Fig. 4A,B). Interestingly, in the peripheral adhesions only approximately 30 and 50 percent of integrin α6β4 was found to be associated with BP230 and keratin, respectively (Fig. 4B). The possibility that the peripheral adhesions lacking keratin and BP230 are associated with the IF protein vimentin can be largely excluded since few PA-JEB/β4 keratinocytes express vimentin and even fewer cells have it associated with peripheral adhesions. For similar reasons, we can exclude that the central adhesions are associated with vimentin (Fig. S1). F-actin, which in many PA-JEB/β4 keratinocytes is organized in a typical circumferential ring, was not obviously associated with either the central or peripheral adhesions (Fig. S1). These results characterize many of the peripheral but not the central integrin α6β4-containing adhesions as type I HDs. Consistent with the notion that type I HDs are present in CD151-deficient patients and mice, the HD proteins BP180, BP230 and plectin were all found to co-localize with integrin α6β4 in the peripheral adhesions of CD151-deficient PA-JEB/β4 keratinocytes (Fig. S2). The central integrin α6β4-containing adhesion clusters, which share some of the characteristics of HDs, such as the clustering of integrin α6β4 and its association with laminin-332 and plectin, also have features in common with TEMs: they contain integrin α3β1 and CD151 and therefore display a hybrid molecular composition. From this point forward, we will refer to these adhesions as central HD-like adhesions in this report.

Despite the fact that CD151 can form stable complex with both integrins α3β1 and α6β4, type I HDs in the skin contain only complexes of CD151 with integrin α6β4 (Sterk et al., 2002; Underwood et al., 2009). To investigate the distribution of integrin α3β1 in relation to CD151 in the different cell-matrix adhesions in the cultured keratinocytes, we performed TIRF imaging and created intensity profiles along a line segment. As shown in Fig. 4, the intensity plots of CD151 and integrin α6β4 in the central HD-like adhesions and the peripheral HDs almost completely overlapped each other (Fig. 4C). In contrast, the profiles of integrins α3β1 and α6β4 only overlapped in the central HD-like adhesions (Fig. 4C). In the peripheral region the peak intensities of integrin α3β1 overlapped with those of tyrosine phosphorylated FA-associated protein paxillin, and not with HD markers BP230 and keratin (Fig. 4D). These results confirm that central and peripheral adhesions differ in protein composition. Furthermore, while in the central HD-like adhesions integrin α3β1, integrin α6β4 and CD151 cluster together, in the periphery they can be found in two distinct complexes: integrin α3β1 in FAs and integrin α6β4 and CD151 in type I HDs.

Our findings so far indicate that CD151 is essential for maintaining central HD-like adhesions. To determine whether CD151 stabilizes these adhesions through binding to integrin α3β1, we deleted and reconstituted integrin α3 in the HaCaT and PA-JEB/β4 keratinocyte cell lines (Fig. 5A,B; Fig. S3A,B). In both these cell lines, the absence of integrin α3 did not affect the total levels of integrin β4 at the cell surface and the formation of peripheral HDs was not obviously disturbed. However, as observed in the CD151-deficient PA-JEB/β4 keratinocytes, the absence of integrin α3 caused a loss of central HD-like adhesions, albeit less dramatically (Fig. 5B–D; Fig. S3C). To exclude the possibility that the absence of the central HD-like adhesions is caused by reduced spreading of the integrin α3-deficient cells, we also analyzed the cells that were plated on coverslips coated with a mixture of vitronectin and fibronectin. Under these conditions, cell spreading of the integrin α3-deficient cells was increased and found to be comparable to that of the wild-type cells on uncoated coverslips. Importantly, the difference in the amount of central HD-like adhesions persisted between the integrin α3-deficient and -proficient cells (Fig. S3D,E).

Next, we investigated whether the role of integrin α3β1 in maintaining the central HD-like adhesions requires binding of integrin α3β1 to laminin-332. To this end, we introduced a binding-deficient integrin α3 mutant (G163A) into integrin α3-deficient HaCaT cells (Zhang et al., 1999). Expression of the mutant integrin α3β1 was verified by FACS analysis (Fig. 5B). Unlike the wild-type integrin α3 subunit, the integrin α3-G163A mutant was unable to completely rescue the central HD-like adhesions, suggesting that integrin α3β1 binding to laminin contributes to efficient maintenance of these adhesions (Fig. 5C,D). These findings were confirmed in PA-JEB/β4 cells using the integrin α3 blocking antibody, J143 (Fig. S3F,G).

Palmitoylation of CD151 is necessary for stabilization of the CD151 TEMs and hence for incorporation and clustering of integrins α3β1 and α6β4 into these multi-molecular complexes of CD151 (Berditchevski et al., 2002; Yang et al., 2002). To study the role of this self-associating function of CD151 for stabilization of the central HD-like adhesions, we introduced a CD151 palmitoylation mutant into the CD151-deficient PA-JEB/β4 keratinocytes (Fig. 5B) (Berditchevski et al., 2002). The expression of this mutant in CD151 KO cells did not rescue the central HD-like adhesions, suggesting that for the maintenance of these clusters CD151 needs to form multi-molecular complexes by oligomerization (Fig. 5E,F; Fig. S3H).

Taken together, these results show that for efficient maintenance of the central HD-like adhesions, both laminin-332-associated integrin α3β1 and self-associated and TEM-forming CD151 are needed.

Since the peripheral HDs and central HD-like adhesions contain more plectin and laminin, respectively, we studied whether CD151-mediated clustering of integrin α6β4 in the different adhesions requires binding of integrin α6β4 to laminin-332 or plectin. Therefore, we expressed the integrin β4 mutants, β4-AD and β4-RW, in integrin β4-deficient PA-JEB keratinocytes (Fig. 6A). The β4-AD mutant is an adhesion-defective mutant that cannot bind to laminin-332, while the β4-RW mutant, carrying a single missense mutation R1281W in the cytoplasmic domain of integrin β4, is unable to bind plectin (Geerts et al., 1999; Nievers et al., 2000; Russell et al., 2003). In the β4-AD mutant-expressing cells, central HD-like adhesions were less commonly present than in the wild-type integrin β4 (β4-WT)-expressing cells (Fig. 6B,C). Similar findings were obtained when PA-JEB/β4 and HaCaT cells were treated with the integrin β4 blocking antibody ASC8 (Russell et al., 2003). In the presence of this blocking antibody, the central HD-like adhesions are mostly absent, while there are abundant peripheral HDs (Fig. 6D,E). In the cells expressing the β4-RW mutant and in plectin-deficient PA-JEB/β4 keratinocytes, the integrin α6β4 staining was diffused and adhesions were less well-defined. Nevertheless, in many cell islands some central HD-like adhesions could be observed (Fig. 6B,C,F,H). In addition, the loss of plectin resulted in a reduction of peripheral HDs (Fig. 6G,H). These results indicate that the binding of integrin α6β4 to laminin-332, as for integrin α3β1, contributes to the stabilization of the central HD-like adhesions. By contrast, binding of integrin α6β4 to plectin is not essential, but might contribute to the clustering of this integrin in the peripheral type I HDs.

Integrin clustering determines the strength with which cells adhere to the extracellular matrix (ECM) (Puklin-Faucher and Sheetz, 2009). To investigate whether CD151-mediated clustering of integrins α3β1 and α6β4 strengthens the adhesion of keratinocytes to laminin-332 deposited by the cells to the substratum, we quantified the adhesion strength of CD151-proficient and -deficient HaCaT keratinocytes, which express both integrin α3β1 and α6β4, using a spinning disk device (Boettiger, 2007). CD151-deficient cells showed a reduced tolerance (albeit not statistically significant) of shear stress compared to the CD151-proficient cells (Fig. 7A), suggesting that CD151 plays only a minor role in adhesion strengthening when both integrins α3β1 and α6β4 are present. Next, we investigated whether cell adhesion mediated by either integrin α3β1 or α6β4 is strengthened by CD151. To resolve this, we individually deleted integrin α3 and integrin β4 in the CD151-deficient HaCaT keratinocytes and confirmed the absence of these proteins by FACS and WB analysis (Fig. 7B,C). The deletion of integrin α3 in the CD151-deficient HaCaT keratinocytes had no effect on the adhesive strength of integrin α6β4 to laminin-332 (Fig. 7D). However, in the absence of integrin α6β4, the additional loss of CD151 resulted in a significant reduction in integrin α3β1-mediated adhesion strengthening (Fig. 7E,F). We conclude that CD151 significantly strengthens integrin α3β1-, but not integrin α6β4-mediated cell adhesion by facilitating the clustering of this integrin into central HD-like adhesions.

CD151 has been shown to be broadly expressed by a variety of cell types and to form complexes with different laminin-binding integrins, including integrins α3β1 and α6β4 (Sincock et al., 1997). In skin, CD151 is complexed with integrin α6β4 and is a component of HDs (Sterk et al., 2000). In this study, we show that in cultured keratinocytes CD151 is not only present in HDs, but also in an adhesion structure with a distinct molecular composition of a HD with tetraspanin features. These hybrid adhesion structures, which are present in the central region of the cell, and referred to as central HD-like adhesions, contain the cytolinker plectin in addition to the CD151–α3β1/α6β4 integrin complexes, but are not associated with keratin filaments. We further show that CD151, through its binding to integrin α3β1, is critical for the formation and stabilization of these central HD-like adhesions and strengthens integrin α3β1- but not integrin α6β4-mediated keratinocyte adhesion to laminin-332. CD151 is not needed for the formation of type I HDs that are formed in the periphery of the cells and are anchored to keratin filaments (Fig. 8).

The mechanistic reasons for the lack of association of keratin filaments with the central HD-like adhesions can be surmised based on previous studies. PKA-mediated phosphorylation of plectin at S4642 interferes with its interaction with intermediate filaments (Bouameur et al., 2013), and therefore, it is possible that plectin is phosphorylated at this residue in the central HD-like adhesions, which prevents its association with keratin filaments. Additionally, BP230, which is present in peripheral HDs but not central HD-like adhesions, may be required for efficient attachment of keratin filaments to the plasma membrane. In support of this supposition, BP230 knockout mice showed an impaired connection of keratin filaments with HDs (Guo et al., 1995). Similarly, reduced amounts of keratins are associated with type II HDs, which contain plectin but lack BP230 (Fontao et al., 1997; Uematsu et al., 1994).

Although an essential role of CD151 in the formation of type I HDs was not evident from our current experiments, a more subtle role of CD151 for the formation of HDs cannot be excluded. The ability of CD151 to facilitate the clustering of integrins α3β1 and α6β4 into central HD-like adhesions may strengthen keratinocyte adhesion to newly deposited laminin-332 before FAs and HDs are fully established; however, once integrins are connected to the cytoskeleton, the function of CD151 as a clustering molecule may no longer be required. This model would indicate an important role of CD151-mediated adhesion strengthening and clustering of integrins α3β1 and α6β4 into central HD-like adhesions in vivo. For example, during re-epithelization of wounded skin, when HDs are dissolved and stable adhesion of the keratinocytes to the basal lamina is no longer assured, the central HD-like adhesions may prevent detachment of migrating keratinocytes that are subject to external mechanical forces. At the same time, they may render the adhesion of keratinocytes sufficiently dynamic to allow integrin turnover and migration to occur. Consistent with a role of the central HD-like adhesions in migration, we found that these adhesions are more dynamic than the type I HDs.

Despite the fact that the association of BP230 with keratin filaments at type I HDs increases their stability (Michael et al., 2014; Seltmann et al., 2013), our FRAP analyses showed only a modest increase in the stability of the peripheral HD adhesions compared to the central HD-like adhesions. Since our FRAP analyses do not distinguish between peripheral adhesions that are or are not associated with keratin filaments, this apparent discrepancy may be explained by the notion that not all of the peripheral adhesions are associated with keratin filaments. Furthermore, it should be understood that the peripheral type I HDs lack integrin α3β1, whose presence in the HD-like adhesions contributes to their stability.

In central HD-like adhesions, CD151 and integrin α3β1 colocalize, but in the periphery CD151 associates with type I HDs and integrin α3β1 is localized into FAs. The localization of integrin α3β1 into FAs coincides with the incorporation of BP230 in HDs. Since the localization of BP180 in type I HDs precedes that of BP230 (Koster et al., 2003), it is possible that BP180 displaces integrin α3β1 from CD151 and allows the integrin to become incorporated into FAs. In this respect it is interesting to mention that, like CD151 and the integrin α3, α6 and β4 subunits (Yang et al., 2004), BP180 contains several cysteine residues adjacent to its transmembrane domain that are potential sites of palmitoylation. The role of protein palmitoylation in the formation of HDs is not known, but it is reasonable to suggest that it might help to increase the local concentration of HD proteins, and facilitate the assembly and stability of HDs.

In agreement with our finding that HDs can be formed in the absence of integrin α3β1 in cultured keratinocytes, no obvious abnormalities in HDs have been observed in mice lacking integrin α3 in basal keratinocytes (Conti et al., 2003; DiPersio et al., 1997; Margadant et al., 2009). Yet, genetic ablation of the β1 integrin subunit or its co-activator kindlin-1 (encoded by FERMT1) in basal keratinocytes has been associated with reduced HD protein expression and formation of HDs (Brakebusch et al., 2000; Qu et al., 2012; Raghavan et al., 2000). This suggests that if integrin α3β1 has a role in HD formation, this role could be fulfilled by other integrins, as well. Furthermore, our finding that CD151 contributes to integrin α3β1- but not integrin α6β4-mediated strengthening of cell adhesions is consistent with the respective roles these proteins play in mediating cell-matrix adhesion in the skin and kidney. In the skin, where basal keratinocytes express both integrins α3β1 and α6β4, the effect of CD151 deletion is relatively mild, and patients display only a mild skin blistering disorder. In contrast, the loss of CD151 in glomerular podocytes, which express integrin α3β1 but not integrin α6β4, has been associated with progressive and irreversible kidney dysfunction (Crew et al., 2004; Kagan et al., 1988; Sachs et al., 2012; Vahidnezhad et al., 2017).

In conclusion, we show that CD151 stabilizes the central HD-like adhesions by strengthening the adhesion of integrin α3β1 to laminin-332 and facilitating the incorporation of integrin α6β4 into TEMs containing integrin α3β1. By contrast, CD151 is not needed for the formation or stabilization of integrin α6β4-containing type I HDs in the periphery of the cells, which are linked to keratin filaments. These data provide new insights into the role of CD151 in the formation and maintenance of different populations of adhesions that are formed in cultured keratinocytes. Importantly, our study reflects the skin conditions observed in patients and mice with mutations in CD151 and integrin α3β1, and thus provides better understanding of the mechanisms underlying these conditions.

The primary and conjugated antibodies used in this study are listed in Table S1. Secondary antibodies for western blotting were goat anti-mouse IgG HRP (1:3000; Bio-Rad, 170-6516) and polyclonal goat anti-Rabbit IgG HRP (1:5000; Dako, P044801-2). Donkey anti-mouse IgG PE (1:500; Jackson ImmunoResearch, 715-116-150) was the secondary antibody used in FACS. For immunofluorescence (IF), rabbit anti-rat IgG (1:1000; in-house), goat anti-guinea pig IgG Alexa Fluor 488, goat anti-human IgG Alexa Fluor 488, goat anti-mouse Alexa Fluor 568, goat anti-mouse Alexa Fluor 647, goat anti-rabbit Alexa Fluor 488, goat anti-rabbit Alexa Fluor 647, goat anti-rat Alexa Fluor 647 (1:200; Invitrogen, A-11073, A-11013, A-11004, A-21236, A-11034, A-21245, A-21247), goat anti-mouse FITC, goat anti-rat FITC (1:200; Rockland, 610-102-121 and 612-102-120) and the isotype-specific goat anti-mouse IgG1 Alexa Fluor 568 (1:500; Thermo Fisher Scientific, A-21124) and goat anti-mouse IgG2a FITC (1:250; Abcam, ab98697) were used secondary antibodies.

PA-JEB immortalized keratinocytes were isolated from a patient with pyloric atresia associated with junctional epidermolysis bullosa (Niessen et al., 1996; Schaapveld et al., 1998). The derivation of this cell line was done for diagnostic purposes, thus, the research conducted using these cells was exempt from the requirement for ethical approval. PA-JEB/β4 keratinocytes stably expressing wild-type β4 were generated by retroviral transduction (Sterk et al., 2000) and maintained in serum-free keratinocyte medium (KGM; Invitrogen), supplemented with 50 µg ml⁻¹ bovine pituitary gland extract, 5 ng ml⁻¹ EGF, and antibiotics (100 units ml⁻¹ streptomycin and 100 units ml⁻¹ penicillin; Sigma-Aldrich). Human HaCaT keratinocytes were obtained from Dr Fusenig of the German Cancer Center (DKFZ) in Heidelberg, Germany and cultured in Dulbecco's modified Eagle's medium (DMEM; Gibco) containing 10% heat-inactivated fetal calf serum (FCS; Serana) and antibiotics. All cells were cultured at 37°C in a humidified, 5% CO₂ atmosphere.

The integrin α3-, integrin β4-, CD151- and plectin-deficient cells were prepared using CRISPR-Cas9 technology. Target sgRNAs against integrin α3 (5′-GCTACTCGGTCGCCCTCCAT-3′), integrin β4 (5′-GACTCGCTCCCTCCGCGCCT-3′), CD151 (5′-CAGGTTCCGACGCTCCTTGA-3′) and plectin (5′-GAGGTGCTTGTTGACCCACT-3′) were cloned into the pX330-U6-Chimeric_BB-CBh-hSpCas9 (Addgene plasmid #42230, deposited by Feng Zhang). For the integrin α3-, integrin β4- and CD151-deficient cells, cells were transiently transfected with this plasmid using Lipofectamine^® 2000 (Invitrogen) in OptiMEM (Gibco). The cells were bulk sorted for the population deficient of the protein of interest using a Moflo Asterios (Beckman Coulter) or FACSAria Fusion (BD Biosciences) cell sorter. For the preparation of the plectin-deficient cells, cells were co-transfected with a plasmid containing a blasticidin cassette flanked by TIA target sites (described by Blomen et al., 2015), blasticidin-resistant cells were enriched by selection for four days in the presence of blasticidin (4 µg ml⁻¹; Sigma-Aldrich), clones were picked and analyzed for plectin expression by western blotting.

For (re)expression of proteins in protein-deficient cells, we used stable cellular retroviral-mediated transduction. Cells that expressed the protein were enriched by zeocin or blasticidin selection and/or by bulk sorting by FACS. The constructs used were: the β4-AD mutant (D230A, P232A and E233A) in LZRS-IRES-blasticidin, a kind gift from Peter Marinkovich (Russell et al., 2003), the CD151 wild-type and CD151 palmitoylation mutant in LZRS-IRES-Zeo, constructed by introducing the CD151 cDNA released from pZeo-CD151 or pZeo-CD151-PALM [gift from F. Berditchevski (Berditchevski et al., 2002)] and the integrin α3 wild-type in LZRS-IRES-Zeo. The integrin α3 cDNA has been described previously (Delwel et al., 1994) and the full-length cDNA encoding human integrin α3 was released from pUC18 by digesting with SphI, made blunt by Klenow digestion and ligated into the SwaI site of the retroviral LZRS-IRES-zeo vector. Furthermore, the integrin α3-G163A mutant in pBJ-1 expression vector (Zhang et al., 1999) was a gift from Yoshikazu Takada (University of California, Davis, USA) and transiently transfected in α3-deficient cells. Generation of PA-JEB keratinocytes stably expressing integrin β4 [β4RW (R1281W)] has been described previously (Geerts et al., 1999).

For image acquisition, cells were seeded at ∼10–20% confluence onto glass coverslips, uncoated or coated with 10 µg ml⁻¹ fibronectin from bovine plasma (F1141, Sigma-Aldrich) and 5 µg ml⁻¹ vitronectin (SRP3186, Sigma-Aldrich) and grown for 40–44 h of which at least the last 20 h in the presence of DMEM+FCS. After the cells had been fixed with 2% PFA for 10 min, permeabilized with 0.2% Triton X-100 and blocked with 2% BSA (resolved in PBS; Serva), they were incubated for 50–60 min with primary and secondary antibodies, with three PBS washes in between. Nuclei were counterstained with DAPI (Sigma-Aldrich) and coverslips were mounted with MOWIOL for confocal imaging. For TIRF imaging, coverslips were stored in PBS at 4°C until imaging.

Confocal imaging was performed using a Leica TCS SP5 confocal microscope with a 63× (NA 1.4) oil objective. A Leica SR-GSD microscope (Leica Microsystems, Wetzlar, Germany) equipped with a 160× oil immersion objective (NA 1.47) and an EMCCD Andor iXon camera (Andor Technology, Belfast, UK) was used to acquire TIRF images with an imaging depth of 200 nm. To be able to perform subsequent analyses, imaging was performed under consistent conditions.

Image analyses were performed using ImageJ: Co-localization was assessed using Manders’ overlap coefficient (MOC) with the JACoP plugin. The amount of basal cell surface clustering was quantified by calculating the integrated density of the staining. The intensity of the integrin β4 staining in the peripheral adhesions was obtained by drawing a region of interest around the peripheral adhesions and measuring the mean intensity of the integrin β4 staining in this region. The percentage of integrin β4 associated with plectin, BP230, keratin 14 or laminin-332 was obtained by calculating how much of the integrin β4-positive area was additionally positive for plectin, BP230, keratin or laminin-332. The quantification of the island sizes was performed by thresholding images based on their staining with Phalloidin-iFluor647 (AAT Bioquest) and measuring the surface area after selecting the outer region. Furthermore, the quantification of the amount of central integrin β4 clusters present per cell island was performed using an arbitrary scoring system (0=no central adhesions, 1=few central adhesions, 2=several central adhesions, 3=many central adhesions).

Graphs were made using GraphPad Prism and include bar graphs showing the mean±s.d. (including dots that represent independent measurement) or boxplots dividing the data in quartiles. The number of images used for analyses is specified in the figure legends and statistical significance was calculated using the non-parametric and non-paired Mann–Whitney U-test; P-values: *<0.05, **<0.01, ***<0.001, ****<0.0001.

For live imaging and fluorescence recovery after photobleaching (FRAP), cells were seeded on 24 mm coverslips and imaging was performed using a Leica TCS SP5 confocal microscope with a 63× (NA 1.4) oil objective equipped with a live cell chamber. Live imaging of CD151-proficient and -deficient PA-JEB/β4–GFP cells was started 20 h after seeding and immediately after replacing the KGM medium with DMEM+FCS. Every hour the differential interference contrast (DIC) and GFP fluorescence signals were collected from multiple cell islands at different positions. Videos were made in ImageJ.

FRAP experiments were performed, using the FRAP wizard in LAS-AF (Leica), using PA-JEB/β4–GFP cells grown for 20 h in KGM medium and subsequently grown for an additional 20 h in DMEM+FCS. Simultaneously, a region in one peripheral and one central integrin α6β4-containing adhesion was bleached and fluorescent recovery was followed over time. Two images as controls were taken before bleaching and then one image every 10 s (first 2 min) or 30 s for 15 min after bleaching. For analyses, the background was subtracted from the measured intensities and the curves were normalized to the mean value before bleaching. Furthermore, the values of the first 230 s were fitted with a single exponential decay using MATLAB software and the fitted curve was used to calculate the mobile fraction and the recovery time τ. Graphs were made using GraphPad Prism; the line plot shows the mean±s.d. and the boxplots divide the data in quartiles. The number of images used for analyses is specified in the figure legends and the significance difference between the central and peripheral adhesions was calculated using the non-parametric Wilcoxon matched-pairs signed rank test (at different time points); P-values: **<0.01, ***<0.001.

For FACS analyses, cells were collected after trypsinization, and incubated for 50–60 min with primary and secondary antibodies on ice in PBS containing 2% FCS. In between the antibody steps, cells were washed twice with 2% FCS in PBS. Finally, cells were passed through a nylon mesh filter and 50000+ cells were analyzed per sample using a FACSCalibur cell analyzer (BD Biosciences). Cells incubated with only secondary antibody were used as a negative control. Graphs include values of a representative experiment (out of three repetitions) normalized to the mode. The graphs were made in FlowJo and adapted in Adobe Illustrator.

To measure long-term adhesion strength of cells, we used the spinning disk assay described by Boettiger (2007). Wild-type or KO HaCaT cells were grown to 60–100% confluence on 30 mm coverslips for 2 days. Just before spinning, the cells were washed briefly with PBS and the coverslips were ‘glued’ to the stamp using grease. The cells were spun for 8 min at 4240 rpm in PBS-ABC buffer [80% A (10 gl⁻¹ NaCl, 0.25 gl⁻¹ KCl, 0.95 gl⁻¹ Na₂HPO₄*2H₂, 0.25 gl⁻¹ KH₂PO₄), 10% B (1.34 gl⁻¹ CaCl₂.2H₂O) and 10% C (2 gl⁻¹ MgCl₂.6H₂O)] containing 5% dextran from Leuconostoc spp. M_r 450,000–600,000 (Sigma-Aldrich). After spinning, the cells were fixed in 2% PFA, permeabilized in 0.2% Triton X-100, stained with DAPI and mounted on microscopic slides using MOWIOL. The complete coverslips were imaged using a Zeiss AxioObserver Z1. The coverslip area was divided in rings with a 1.5 mm diameter and the cell confluence was measured per ring and normalized to the total area of the ring and to the confluence in the middle area (100% confluent). The average shear stress per ring was calculated using the formula with τ is the shear stress in dynes/cm², r is the radial distance from the center of rotation in m, ρ is the fluid density in kg m⁻³, μ is the fluid viscosity in Pa s and ω is the rotational velocity in radians min⁻¹. The τ50 was obtained by fitting the data points to a logistic function using MATLAB software, where c is the inflection point and equal to the point of 50% cell detachment (τ50) and b is the slope at the inflection point.

Graphs were generated and significance was calculated using GraphPad Prism. The line plots show the mean value of the 9–16 experiments per condition, which are individually shown as squares, diamonds or triangles. For the wild-type line, the same dataset was used in the different graphs. The significance of the percentage of adherent cells in KO cells compared to the wild-type cells was calculated at each data point using unpaired t-tests corrected for multiple comparisons using the Holm–Sidak method; P-values: ***<0.001, ****<0.0001. The τ50 was visualized in a bar graph showing the mean±s.d. of 9–10 independent experiments. The significant different between the τ50 in integrin β4 KO and integrin β4+CD151 KO cells was calculated using the non-parametric Mann–Whitney U-test; P-value: ****<0.0001.

For analysis of proteins in whole cell lysates, sub-confluent cells were washed in cold PBS, lysed in RIPA (1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 4 mM EDTA pH 7.5, 100 mM NaCl, 20 mM Tris-HCl pH 7.5) supplemented with inhibitors (1.5 mM Na₃VO₄, 15 mM NaF and 1:500 protease inhibitor cocktail), whole-cell lysates were cleared by centrifugation at 15,000 g for 60 min at 4°C and supplemented with SDS sample buffer (50 mM Tris-HCl pH 6.8, 2% SDS, 10% glycerol, 12.5 mM EDTA, 0.02% Bromophenol Blue) with or without β-mercaptoethanol and heated at 95°C for 5 min. Immunoprecipitations were performed with cell lysates prepared in 1% NonidetP-40, 100 mM NaCl, 4 mM EDTA, 20 mM Tris-HCl (pH 7.5), supplemented with inhibitors. After centrifugation at 15,000 g for 60 min at 4°C, the cleared lysates were incubated at 4°C for 1.5 h with 1 µg ml⁻¹ antibody. Subsequently, the lysates were incubated overnight with Protein G Sepharose 4 Fast Flow beads (GE Healthcare), beads were washed three times with lysis buffer and two times with PBS and bound proteins were dissolved in SDS sample buffer without β-mercaptoethanol and heated at 95°C for 5 min.

Proteins were separated on 4–12% bolt Novex gradient gels (Invitrogen) and transferred to Immobilon-P transfer membranes (Millipore). The membrane was blocked for at least 2.5 h in 2% BSA in TBST (10 mM Tris pH 7.5, 150 mM NaCl, 0.05% Tween 20) before incubation with primary antibody overnight at 4°C and with secondary antibody for 1 h at room temperature. After each incubation, membranes were washed twice with TBST and twice with TBS (TBST without Tween 20). Antibodies were detected using Clarity Western ECL Substrate (Bio-Rad) and in the figures the molecular mass markers (kDa) are indicated.

We thank Leila Nahidiazar, Bram van den Broek, Lenny Brocks and Marjolijn Mertz for their help with imaging and image analysis; Martijn van Baalen, Anita Pfauth, Frank van Diepen and Debajit Bhowmick for their help with cell sorting; Alba Zuidema, Veronika Ramovs, Jose de Pereda and Kevin Wilhelmsen for critical reading of the manuscript. Furthermore, we thank our colleagues for sharing reagents.