Functional differences between kindlin-1 and kindlin-2 in keratinocytes

Abstract

Integrin-β1-null keratinocytes can adhere to fibronectin through integrin αvβ6, but form large peripheral focal adhesions and exhibit defective cell spreading. Here we report that, in addition to the reduced avidity of αvβ6 integrin binding to fibronectin, the inability of integrin β6 to efficiently bind and recruit kindlin-2 to focal adhesions directly contributes to these phenotypes. Kindlins regulate integrins through direct interactions with the integrin-β cytoplasmic tail and keratinocytes express kindlin-1 and kindlin-2. Notably, although both kindlins localize to focal adhesions in wild-type cells, only kindlin-1 localizes to the integrin-β6-rich adhesions of integrin-β1-null cells. Rescue of these cells with wild-type and chimeric integrin constructs revealed a correlation between kindlin-2 recruitment and cell spreading. Furthermore, despite the presence of kindlin-1, knockdown of kindlin-2 in wild-type keratinocytes impaired cell spreading. Our data reveal unexpected functional consequences of differences in the association of two homologous kindlin isoforms with two closely related integrins, and suggest that despite their similarities, different kindlins are likely to have unique functions.

Introduction

Integrins are a large family of αβ heterodimeric cell adhesion receptors with vital roles in cell adhesion, assembly of the extracellular matrix and intracellular signaling (Hynes, 2002). Integrins interact with extracellular ligands, such as fibronectin, laminin and collagens, through their large ectodomains, and couple to the cytoskeleton and diverse intracellular signaling networks through the assembly of multi-protein complexes that are linked to the short cytoplasmic tails of the integrins (Harburger and Calderwood, 2009). Many integrin-associated proteins are known (Zaidel-Bar et al., 2007), and key common components of the cytoplasmic complexes, such as the integrin-activating protein talin, have been identified. However, although the extracellular ligand specificity of different integrin heterodimers is well established (Humphries et al., 2006), the basis for differential intracellular signaling and cytoskeletal responses following ligation of different classes of integrins is less well understood.

Despite extensive conservation between different integrin subunits and some partial redundancy in function, β1 integrins play unique essential roles in the assembly and organization of the extracellular matrix and in intracellular signaling. This is clearly evident in the epidermis, where, despite compensatory upregulation of the β6 integrin subunit, conditional ablation of β1 integrin results in severe blistering and basement membrane defects (Brakebusch et al., 2000; Raghavan et al., 2000). In culture, keratinocytes adhere to the underlying fibronectin-rich substratum through α5β1 and αvβ6 integrins (Watt, 2002). Integrin-β1-knockout (β1-KO) keratinocytes can adhere to fibronectin as a result of upregulation of αvβ6 integrins, but they display altered adhesion, cell spreading and polarized migration in vitro (Raghavan et al., 2003). That these differences are evident even on experimentally defined substrates, which bind both β1 and β6 integrins, suggest that the ability of β1 and β6 integrins to recruit distinct intracellular signaling and adaptor proteins accounts for some of the differences between them.

Recent work has highlighted the importance of kindlins for the normal functioning of integrins (Larjava et al., 2008; Meves et al., 2009). Kindlins are FERM-domain-containing proteins that directly bind to the membrane-distal NPxY motifs within the β integrin tail. Only two of the three mammalian kindlins are expressed in keratinocytes: the epithelial-specific kindlin-1 and the essential ubiquitously expressed kindlin-2 (Lai-Cheong et al., 2008). Mutations in the kindlin-1 gene cause Kindler's syndrome, a rare human disease that results in skin blistering and atrophy (Ashton et al., 2004; Siegel et al., 2003). Likewise, kindlin-1-KO mice display reduced keratinocyte proliferation and some skin atrophy (Ussar et al., 2008), although they do not develop skin blistering, and their basement membrane is normal, suggesting that kindlin-2 partially compensates in the kindlin-1-KO skin. Consistent with this, both kindlin-1 and kindlin-2 localize to the focal adhesions in keratinocytes and have overlapping functions in human keratinocytes (He et al., 2011a). Nonetheless, the fact that endogenous kindlin-2 is unable to fully compensate for kindlin-1 deficiency in human disease or mouse knockouts, suggests that kindlins might have functionally significant, as yet uncharacterized, differences in their interaction partners. Here we show that although kindlin-1 and kindlin-2 both bind and colocalize with β1 integrins, only kindlin-1 binds β6 integrins, and consequently, kindlin-2 is not recruited to the integrin-β6-rich adhesions in β1-KO keratinocytes. We further show that, even in the presence of normal kindlin-1 recruitment, defective kindlin-2 binding results in defective keratinocyte spreading. Taken together, our data point to important functional differences between kindlin-1 and kindlin-2 in keratinocytes.

Results

Localization of integrin β1 cytoplasmic-domain-interacting proteins in β1-KO cells and skin

β1-KO cells can adhere to fibronectin through integrin αvβ6; however, the KO cells have large peripheral focal adhesions and are defective in cell spreading. To better understand the differences between the focal adhesions nucleated by integrin β1 and β6 on a defined substratum (fibronectin), we looked to see whether there were any differences in recruitment of key β1-tail-associated proteins (vinculin, paxillin, FAK, talin, kindlin-1, kindlin-2 and ILK) to the focal adhesions of wild-type (WT) and β1-KO cells. In WT cells, both β1 and β6 integrins were localized to peripheral focal adhesions (Fig. 1A,A′), whereas the smaller central focal complexes were primarily nucleated by β1 integrin (arrowheads, Fig. 1A). β1-KO cells were less well spread than the WT controls but formed large peripheral focal adhesions that were nucleated by integrin β6 (Fig. 1B,B′). In WT keratinocytes, vinculin, paxillin and FAK were localized to the β1-containing focal adhesions and complexes (Fig. 1C,C′,E,E′,G,G′). Interestingly there was robust recruitment of vinculin, paxillin and FAK to the focal adhesions of KO keratinocytes (Fig. 1D,D′,F,F′,H,H′). Next, we looked at the recruitment of talin, kindlin-1, kindlin-2 and ILK in WT and KO focal adhesions. In both WT and KO cells we observed robust recruitment of talin (Fig. 2A–B′) and kindlin-1 (Fig. 2C–D′) to the focal adhesions. However, whereas kindlin-2 and ILK were strongly recruited to the focal adhesions of WT cells (Fig. 2E,E′,G,G′), there was very poor recruitment of kindlin-2 to the β1-KO focal adhesions, and instead, its expression was diffuse throughout the cell (Fig. 2F, arrowheads in 2F′). ILK targeting was also perturbed in the β1-KO cells because we observed only weak ILK recruitment to the β6-mediated focal adhesions (Fig. 2H,H′). However, the ILK targeting defect appeared less severe than that of kindlin-2 (compare Fig. 2F′,H′). Western blot analysis of lysates from WT and KO keratinocytes revealed that there was no significant reduction in the overall expression of these proteins in the KO cells compared with the WT cells (Fig. 2I), suggesting that the loss of integrin β1 did not directly impact the levels of kindlin-2 and ILK, but rather affected their localization to the focal adhesions. Next, we analyzed the expression patterns of these proteins in E17.5 WT and β1-KO skin epidermis. As expected, in the WT skin sections, β1 was expressed both in the epidermis and the dermis, and the basement membrane (BM) was marked with laminin-5 (Fig. 3A,A′). In the conditional KO skin sections, β1 was not expressed in the epidermis (asterisk in Fig. 3B′), whereas the dermal expression was still strong and the BM quite disrupted, as judged by laminin-5 staining (Fig. 3B). Interestingly we found de novo expression of β6 in the KO epidermis (Fig. 3D,D′) compared with the WT epidermis (Fig. 3C,C′). In the WT and β1-KO epidermis, we found normal expression of kindlin-1 in the basal layer (Fig. 3E–F′), whereas kindlin-2, ILK and migfilin (which is recruited by kindlin-2) were not localized normally in the basal keratinocytes of the KO skin (Fig. 3H,J,L; supplementary material Fig. S1B; asterisk in Fig. 3H′,J′,L′; supplementary material Fig. S1B′) compared with the WT epidermis (Fig. 3G,G′,I,I′,K,K′). Taken together, these data suggest that the recruitment of kindlin-2 and ILK, but not talin or kindlin-1, to focal adhesions depends crucially on the presence of integrin β1 and that upregulation of β6 integrin cannot fully compensate for the lack of β1 integrin.

Fig. 1.

Localization of focal adhesion proteins in WT and KO cells. (A–B′) WT (A,A′) and KO (B,B′) cells stained for β1 (green) and β6 (red), showing localization to the peripheral focal adhesions. The smaller focal complexes primarily contain β1 integrin (arrowheads, A). In all panels, the focal adhesions in the WT cells are stained with β1 (green), whereas those in the KO are stained with β6 (green) and co-stained in the WT cells with vinculin (C,C′) and), paxillin (E,E′ and FAK (G,G′) in red and in the KO cells with vinculin (D,D′) paxillin and FAK (H,H′) also in red. Scale bar: 10 μm.

Fig. 2.

Localization of β1-cytoplasmic-domain-interacting proteins in WT and KO cells. In all panels, WT focal adhesions are stained with β1 (green), whereas KO focal adhesions are stained with β6 (green) and co-stained in red with talin (A,B), kindlin-1 (C,D), kindlin-2 (E,F) and ILK (G,H). Talin and kindlin-1 strongly localize to the focal adhesions of both cell types (A′–D′). Kindlin-2 fails to localize to the focal adhesions in the KO cell (arrowheads in F′) and there is very poor recruitment of ILK to the focal adhesion in the KO cells (H′). (I) Western blot analysis of WT and KO lysates with talin, kindlin-1, kindlin-2 and ILK. Tubulin was detected as a loading control. Scale bar: 10 μm.

Fig. 3.

Localization of β1-cytoplasmic-domain-interacting proteins in WT and KO skin. (A) WT and (B) KO skin sections stained with β1 (green) and laminin 5 (red). β1 is localized to the basal cells in WT (A′) and absent from the KO epidermis (B′). (C,C′) WT and (D,D′) KO skin sections stained with β6 (green) and K14 (red). There is de novo expression of β6 in the KO epidermis. (E) WT and (F) KO skin sections stained with kindlin-1 (green) and α6 integrin (red). Kindlin-1 is localized to the basal cells in WT (E′) and KO epidermis (F′). (G) WT and (H) KO skin sections stained with kindlin-2 (green) and laminin 5 (red). Kindlin-2 is not localized to the basal cells in the KO epidermis (asterisks in H′). (I) WT and (J) KO skin sections stained with ILK (green) and laminin 5 (red). ILK is not localized to the basal cells in the KO epidermis (asterisks in J′). (G) WT and (H) KO skin sections stained with migfilin (green) and α6 integrin (red). Migfilin is not localized to the basal cells in the KO epidermis (asterisks in L′). In all the WT and KO panels, the BM is demarcated with a dashed line at the dermal-–epidermal junction. The insets are higher magnification images of the dashed box. Scale bar: 200 μm (A–L); 100 μm (A′–L′). epi, epidermis; de, dermis.

Interaction of β1 and β6 integrin tails with kindlin-1 and kindlin-2

The preceding data indicate that talin and kindlin-1 localize in β6-rich focal adhesions, but that kindlin-2 targets poorly to focal adhesions in β1-KO cells. Talin directly binds the membrane-proximal NPxY motif of most short integrin β-tails whereas kindlin-1 and kindlin-2 have been shown to bind the distal NPxY motifs of integrin β1- and β3-tails (Harburger et al., 2009; Montanez et al., 2008). The β1- and β6-tails are well conserved over much of their length, including key talin-binding residues and both NPxY motifs; however, β6 integrin contains an 11 amino acid extension at the C-terminus (Fig. 4A). To compare β1- and β6-tail interactions with kindlin-1 and kindlin-2, we set up pull-down assays with GST-tagged, recombinant, purified cytoplasmic tails of integrin β1 and β6. First we performed endogenous pull downs using WT mouse keratinocyte lysates using GSTβ1, GSTβ6, GSTβ6Δ (cytoplasmic domain of β6 with the last 11 amino acids deleted, this version of β6 is the same length as β1) and GST as bait. These were probed with antibodies against kindlin-1 and kindlin-2. Our data showed that although GSTβ1 could efficiently pull down kindlin-1 and kindlin2, GSTβ6 could pull down kindlin-1, but not kindlin-2 (Fig. 4B,C). Likewise GSTβ6Δ could pull down kindlin-1 but not kindlin-2, suggesting that the 11 amino acid tail did not alter the interaction of β6 integrin with kindlin-1 and kindlin-2. We next performed pull-down assays with FLAG-tagged kindlin-1 and kindlin-2 expressed in CHO-A5 cells. Similar to the results in the endogenous pull-down experiments, we found that FLAG–kindlin-1 was efficiently pulled down by GSTβ1, GSTβ6 and GSTβ6Δ (Fig. 4D), whereas FLAG–kindlin-2 was only efficiently pulled down by GST–β1 (Fig. 4E). The specificity of the β1–kindlin-2 interactions was confirmed with point-mutated β-tails. Previous work has shown that tyrosine to alanine mutations at position 795 of β1 [β1(Y795A)] blocked binding to kindlin-2, whereas mutating the lysine at position 794 [β1K(794A)] had no effect on kindlin-2 binding (Harburger et al., 2009). Interestingly, the amount of kindlin-2 that was pulled down by GST–β6 and GST–β6Δ was only marginally greater than that binding to GST–β1(Y795A) (Fig. 4E,K). Not surprisingly GST–β1(K794A) was able to pull-down kindlin-2 in a similar manner to GST–β1. Thus, kindlin-2 binds specifically to β1-tails but, consistent with its lack of enrichment at β6-mediated focal adhesions, kindlin-2 binds very poorly to β6 integrin. This effect is kindlin-2 specific because kindlin-1 bound both β1- and β6-tails and targets to β1- or β6-rich adhesions.

Fig. 4.

Mapping the interaction between β-tails and kindlins using GST pull-down assays. (A) Sequences of the β1 and β6 cytoplasmic tails used to generate the GST constructs. Pull-down of endogenous kindlin-1 (B) and kindlin-2 (C) with GST–β1, GST–β6, GST–β6Δ and GST alone. Pull-down of FLAG-tagged kindlin-1 with GST–β1, GST–β6, GST–β6Δ and GST (D) and FLAG-tagged kindlin-2 with GST–β1, GST–β6, GST–β6Δ, GST–β1(K974A), GST–β1(Y795A) and GST (E) expressed in CHOA5 cells. Pull down of FLAG tagged kindlin-1 (F) and FLAG-tagged kindlin-2 (G) with GST–β1, GST–β6, GST–β1-β6, GST–β6-β1 and GST. (F). Pull down of FLAG tagged kindlin-1 (H) and FLAG-tagged kindlin-2 (I) with GST–β1, GST–β6Δ, GST–β6Δβ1(NPKY), GST–β6Δβ1(KSA) and GST. (F). Quantification of the amount of FLAG-tagged kindlin-1 (J) and kindlin-2 (K) pulled down by the various GST constructs. Results are normalized for loading against GST control (NIH ImageJ software). Data show means ± s.e. from three independent experiments.

Our data suggest that kindlin-2 binds specifically to β1 but not β6. To localize the region of the β-tails responsible for this effect, we generated chimeric β1-β6 tails by swapping the sequences of β1 and β6 after the first NPxY motif. GST–β1-β6 was generated by replacing residues 784–798 of β1 with residues 763–788 of β6. GST–β6-β1 was generated by replacing residues 763–788 of β6 with residues 784–798 of β1 (Fig. 4A). We checked for the ability of these GST–β1-β6 and GST–β6-β1 tails to pull down kindlin-1 and kindlin-2. Interestingly, the GST–β1-β6 construct was no longer able to pull down kindlin-2, whereas the GSTβ6-β1 construct was able to pull down kindlin-2 as efficiently as GST–β1 was (Fig. 4G). Consistent with its ability to bind both β1- and β6-tails, the binding of kindlin-1 to GST–β6-β1 and GST–β1-β6 was not affected (Fig. 4F). To further narrow down the sequences responsible for this differential recruitment of the kindlins, we generated two more constructs (Fig. 4A). GST–β6Δ-β1(NPKY) was generated by replacing residues 771–777 of β6Δ with residues 792–798 of β1 and GST–β6Δ-β1(KSA) was generated by replacing residues 763–770 of β6Δ with residues 784–791 of β1. Notably, the GST–β6Δ-β1(NPKY) construct in which the second NPxY motif of β6Δ was replaced with the one from β1 was able to pull down kindlin-2, whereas the GST–β6Δ-β1(KSA) construct was not (Fig. 4I). As expected, kindlin-1 was efficiently pulled down by both GST–β6Δ-β1(NPKY) and GST–β6Δ-β1(KSA) (Fig. 4H). Finally we quantified the amount of kindlin-1 and kindlin-2 that was pulled down by the various GST constructs by averaging the data from three independent experiments. Kindlin-1 was pulled down efficiently by GST–β1, GST–β6, GST–β6Δ, GST–β1-β6, GST–β6-β1 GST–β6Δ-β1(NPKY) and GST–β6Δ-β1(KSA), with virtually no binding seen for GST alone (Fig. 4J). Although kindlin-2 was pulled down efficiently by GST–β1, GST–β1(K794A) GST–β6-β1 and GST–β6Δ-β1(NPKY), the ability of GST-β6 and GST-β6Δ to bind to kindlin-2 was reduced by 80% and that of GST–β1-β6 and GST–β6Δ-β1(KSA) was reduced by 65% compared with levels obtained with GST–β1. As expected, virtually no binding to kindlin-2 was observed by the GST–β1(Y795A) and GST alone constructs (Fig. 4K). Taken together, our data suggest that by swapping small regions of the cytoplasmic tail between β1 and β6, we are able to generate a version of β1 that no longer efficiently recruits kindlin-2 and more importantly, by swapping the NPxY domain of β1 with that of β6, we were able to generate a GST–β6 construct that can pull down kindlin-2 quite efficiently. These data further show, for the first time, that different integrins support differential binding and targeting of kindlin-1 and kindlin-2 within the same cell, thereby revealing potentially important functional differences between kindlin-1 and kindlin-2.

Expression of kindlin-2 is important for cell spreading in keratinocytes

We have previously shown that β1-KO keratinocytes exhibit defective cell spreading (Raghavan et al., 2003), and have now found that they do not recruit kindlin-2 to focal adhesions (Fig. 2). To determine whether kindlin-2 is important for keratinocyte spreading, we knocked down kindlin-2 expression in WT keratinocytes and assessed cell spreading. For the knock down, we purchased the TRC library based lentiviral shRNA clone for kindlin-2 and a scrambled probe from Sigma. WT cells were infected with lentiviruses expressing kindlin-2 shRNA and scrambled shRNA. The infected cells were easily identified because they were resistant to puromycin. Western blot analysis of the lysates obtained from scrambled and kindlin-2-knockdown cells showed that kindlin-2 was knocked down by 85%. As expected, the level of kindlin-1 was similar between the scrambled and kindlin-2-knockdown cells (Fig. 5A). Next, we assessed the ability of kindlin-2-knockdown cells to spread. We measured the area of scrambled and kindlin-2-knockdown cells (Fig. 5B, n=100). The average area of the scrambled knockdown cells was 3183 μm², whereas the area of the kindlin-2-knockdown cells was 2214 μm². The reduction in cell spreading seen in kindlin-2-knockdown cells is in agreement with recent data on spreading of kindlin-2-knockdown fibroblasts (He et al., 2011b). We next looked at the localization of β1 integrin and kindlin-2 in scrambled and kindlin-2-knockdown cells. The scrambled knockdown cells displayed robust expression of β1 and kindlin-2 in the focal adhesions and complexes and were well spread (Fig. 5C,C′). Approximately 5-10% of the knockdown cells had some expression of kindlin-2 but a majority of the knockdown cells did not have any detectable expression of kindlin-2 (Fig. 5D′), but did show localization of β1 to the focal adhesions (Fig. 5D). We next looked at the localization of kindlin-1, β6 integrin and vinculin in the focal adhesions of the knockdown cells. Both β6 and kindlin-1 were associated with the peripheral focal adhesions in the scrambled (Fig. 5E,E′) and kindlin-2-knockdown cells (Fig. 5F,F′). Vinculin was associated with the peripheral focal adhesions and the smaller central focal complexes of scrambled knockdown cells (Fig. 5G,G′). Interestingly, in knockdown cells that were not well spread, the focal adhesions appeared to be larger, and more peripheral, with the smaller central complexes not as visible (Fig. 5H,H′). This phenotype was reminiscent of the β1-KO cells, and was seen in approximately 50–60% of kindlin-2-knockdown cells. Thus, despite the presence of wild-type levels of kindlin-1, and effective focal adhesion targeting of kindlin-1, loss of kindlin-2 impairs keratinocyte spreading. Taken together with the defective kindlin-2 recruitment in β1-KO cells, these data suggest that kindlin-2 localization to adhesions might be required for normal spreading.

Fig. 5.

Expression of kindlin-2 is important for cell spreading in keratinocytes. (A) Western blot analysis of lysates from cells expressing scrambled and kindlin-2 shRNA with kindlin-2 and kindlin-1. Tubulin was detected as a loading control. (B) Box plot of the distribution of the average cell area along the median (n=100) of scrambled and kindlin-2-knockdown cells (***P<0.001, compared with scrambled cells). Scrambled (C,C′) and kindlin-2-knockdown cells (D,D′) stained with β1 (green) and kindlin-2 (red). Scrambled (E,E′) and kindlin-2-knockdown cells (F,F′) stained with β6 (green) and kindlin-1 (red). Scrambled (G,G′) and kindlin-2-knockdown cells (H,H′) stained with vinculin (green) and phalloidin (red). Scale bar: 10 μm.

Role of kindlin-2 recruitment in integrin-β1-mediated cell spreading and focal complex formation

To investigate whether loss of kindlin-2 recruitment contributes to the altered focal adhesion assembly and defective cell spreading seen in β1-KO keratinocytes we generated full-length lentiviral rescue constructs that encode the extracellular and transmembrane domain of β1, in conjunction with the cytoplasmic domain constructs that we used for the GST pull-down assays. We generated (1) β1 (full-length β1), (2) β1-β6 (extracellular and transmembrane domain of β1 with cytoplasmic tail of β6), (3) β1–β1-β6 (extracellular and transmembrane domain of β1 with residues 752-782 of the β1 cytoplasmic tail followed by residues 766–788 of the β6 cytoplasmic tail) and (4) β1–β6-β1 (extracellular and transmembrane domain of β1 with residues 731–762 of the β6 cytoplasmic tail followed by residues 783–798 of the β1 cytoplasmic tail). β1-KO cells were infected with the various rescue constructs and infected cells were sorted by FACS to identify cell populations that expressed WT levels of the chimeric rescue construct as judged by the surface β1 expression. These cells were plated on fibronectin-coated coverslips and analyzed for the expression of β1 and vinculin and β1 and kindlin-2. Two independent KO cell lines were generated and analyzed with each rescue construct. WT cells were well spread and displayed robust β1 integrin staining in focal adhesions and contacts (Fig. 6A; supplementary material Fig. S2A) where it colocalized with vinculin (Fig. 6B; supplementary material Fig. S2B) and kindlin-2 (Fig. 6C,C′; supplementary material Fig. S2C,C′). As expected, β1-KO cells did not have surface β1 and were poorly spread (Fig. 6D; supplementary material Fig. S2D). KO cells displayed robust peripheral focal adhesions that were positive for vinculin (Fig. 6E; supplementary material Fig. S2E) and β6 integrin (Fig. 6F; supplementary material Fig. S2F), whereas kindlin-2 was cytoplasmic and absent from the large β6-rich peripheral focal adhesions (Fig. 6F′; supplementary material Fig. S2F′). In KO cells rescued with the WT β1 construct, approximately 13% of cells expressed WT levels of β1 on their surface (Fig. 6G). Re-expression of β1 in KO cells restored β1 integrin staining at focal adhesions and complexes, and enhanced cell spreading (Fig. 6H; supplementary material Fig. S2H). Kindlin-2 recruitment to focal adhesions in β1-re-expressing cells was comparable to that in WT cells (Fig. 6I,I′; supplementary material Fig. S2I,I′). These data confirm that the recruitment of kindlin-2 to the focal adhesions is crucially dependent on the expression of β1 integrin. In KO cells rescued with the β1-β6 construct, approximately 18% of cells expressed high levels of the chimeric construct on the cell surface (Fig. 6J). Although this chimeric protein was targeted to focal adhesions (as judged by the co-expression of β1 and vinculin), it was unable to restore cell spreading to WT levels (Fig. 6K; supplementary material Fig. S2K). Interestingly, consistent with the inability of β6-tails to efficiently bind kindlin-2, kindlin-2 was not localized to the focal adhesions that were nucleated by β1-β6 (Fig. 6L,L′; supplementary material Fig. S2L,L′). These data suggest that the ability of integrin β1 to recruit kindlin-2 to focal adhesions is a function of sequences in its cytoplasmic domain. In KO cells rescued with the β1–β1-β6 construct, approximately 7% of the cells expressed high levels of cell-surface chimeric integrin (Fig. 6M). In these cells, although this chimeric construct was targeted to the focal adhesions (Fig. 6N; supplementary material Fig. S2N), cell spreading was not rescued. In addition, this construct was not able to recruit kindlin-2 to the focal adhesions (Fig. 6O,O′; supplementary material Fig. S2O,O′). These data suggest that by substituting a small part of the β1-tail with β6, we were able to generate a version of β1 that was no longer able to recruit kindlin-2 nor effectively rescue the β1-KO phenotype. In KO cells rescued with the β1–β6-β1 construct, approximately 10% of the cells expressed high levels of the cell-surface chimeric protein (Fig. 6P). In these cells, the chimeric integrin was targeted to focal adhesions and complexes (Fig. 6Q; supplementary material Fig. S2Q). Interestingly, this construct was able to restore the cell spreading as well as the recruitment of kindlin-2 to the focal adhesions (Fig. 6R′; supplementary material Fig. S2R,R′). These data suggest that specific sequences up to the first NPxY motif of β1 integrin (residues 752–782) are not important for the recruitment of kindlin-2, but that residues 783–798 following the NPxY motif are crucial for kindlin-2 recruitment. This is consistent with the known binding site for kindlins in integrin β-tails. Furthermore, our data suggest that kindlin-2 might be an important mediator of integrin-β1-dependent cell spreading and focal complex formation.

Fig. 6.

Rescue of β1-KO cell spreading correlates with targeting of kindlin-2 to focal adhesions. Surface expression of endogenous β1 on WT (A) and KO (D) cells assessed by FACS. (B,C) WT cells stained for β1, vinculin and kindlin-2, as indicated. (E,F) KO cells stained for β1, vinculin, β6 and kindlin-2, as indicated. (G–R′) Surface expression of β1 on KO cells rescued with a full-length β1 (G), β1-β6 (J), β1–β1-β6 (M), or β1–β6-β1 (P) was assessed by FACS. Cells rescued with full-length β1 (H,I), β1-β6 (K,L), β1–β1-β6 (N,O), or β1–β6-β1 (Q,R) were stained for β1, vinculin and kindlin-2, as indicated. Scale bar: 10 μm.

Cell surface area and focal adhesion size are well correlated with the ability to recruit kindlin-2

Finally, to see whether the rescue of cell spreading correlated with the recruitment of kindlin-2 to focal adhesions and loss of large peripheral adhesions, we quantified the cell area and the size of focal adhesions (assessed by the expression of vinculin in the focal adhesions) of WT and KO cells and then compared it with the area and focal adhesion size of cells that were rescued with β1 integrin (Fig. 7A,B,C). The average area of the WT cells was ~3000 μm² and 100% of these cells exhibited kindlin-2 localization in focal adhesions. The average area of the β1-KO cells was only ~950 μm² and only 5% of these cells had detectable kindlin-2 localization in focal adhesions. However, the average surface area of the focal adhesions in KO cells was 3.8±2.1 μm² (n=1013), which was almost three times the area of adhesions in WT cells 1.3±0.6 μm² (n=1043). The KO cells had a preponderance of large peripheral focal adhesions and virtually no small focal complexes, which probably accounts for the lack of cell spreading. In cells that were rescued with full-length β1, spreading was restored and kindlin-2 strongly localized to focal adhesions in 63% of the cells. Furthermore, the average surface area of the focal adhesions in β1-rescued cells was 1.2±0.8 μm² (n=1041), which was comparable to the surface area of wild-type cells. Interestingly, we found that in the cells where β1 was targeted to the focal adhesions, those cells that had strong kindlin-2 staining were better spread and had smaller focal adhesion areas than those that did not have efficient kindling targeting. Together, these data show that β1 integrin is required for efficient keratinocyte spreading and for targeting kindlin-2 to adhesions.

Fig. 7.

Cell surface area and focal adhesion size are well correlated with the ability to recruit kindlin-2. (A) Table showing the mean focal adhesions and cell surface areas of the WT, KO, rescued cells, kindlin-2 KD and scrambled KD cells. (B) Box plot for average cell area (n=100) for WT, KO and rescued cells (***P<0.001, compared with WT cells, ^P>0.2, compared with WT). (C) Box plots for average focal adhesion area for WT, KO, rescued, kindlin-2 and scrambled KD cells (***P<0.001; FA significantly larger than WT).

The preceding data reveal a strong correlation between kindlin-2 targeting to adhesions, cell spreading and a reduction in adhesion size. However, to further test the significance of kindlin-2 recruitment to adhesions and separate this from effects solely due to β1 re-expression, we investigated cell spreading and focal adhesion area in cells rescued with various chimeric β1-β6 integrins. Expression of either the β1-β6 or the β1–β1-β6 construct which do not efficiently recruit kindlin-2 to adhesions failed to fully rescue spreading or completely restore focal adhesion area to wild-type levels (Fig. 7A–C). Nonetheless, both constructs did produce a partial rescue of these phenotypes and weak kindlin-2 targeting to adhesions was seen in some cells expressing β1-β6 and β1–β1-β6 (38% and 44%, respectively). However, in cells that were rescued with the β1–β6-β1 construct which efficiently rescued kindlin-2 localization to focal adhesions resulting in strong adhesion targeting in 68% of the cells, cell spreading and the size of focal adhesions was fully rescued (Fig. 7A,B,C). Thus a complete rescue of cell spreading and focal adhesion size correlates strongly with targeting of kindlin-2 to adhesions.

In support of the importance of kindlin 2 in determining cell spreading, we noted that the area of KO cells re-expressing the kindlin-2-binding deficient chimeric integrins β1-β6 or β1–β1-β6 was comparable to kindlin-2-knockdown cells (~2214 μm²) and the focal adhesion areas of the kindlin-2-knockdown cells was similarly comparable to the kindlin-2-binding-deficient chimeras, and were significantly larger than those seen in the scrambled knockdown cells (Fig. 7A,C). Taken together, these data suggest that localization of kindlin-2 in the focal adhesions controls keratinocyte spreading and that there is a correlation between the levels of kindlin-2 in focal adhesions and the extent of rescue of cell spreading, as well as the size of the focal adhesions, of the β1-KO rescue cells.

Our data also point to kindlin-2-independent effects of β1 integrin on cell spreading. It seems that all the chimeric integrins tested led to at least a partial rescue of cell spreading (Fig. 7B). Thus, expression of chimeric β1-β6 integrins in KO cells produces a partial rescue without effectively targeting kindlin-2 to adhesions. This partial rescue could not be enhanced by kindlin-2 overexpression because the kindlin-2-overexpressing cells had an area of 1820±98 μm², which was comparable to that of control β1-β6 cells (1975±84 μm²). This suggests that the weak β6–kindlin-2 interaction cannot be compensated for by kindlin-2 overexpression. Instead, the extracellular domain on β1 integrin might provide partial rescue of cell-spreading defects because of a higher avidity for fibronectin. Consistent with this, adhesion assays show that wild-type cells adhere more efficiently than KO cells to a range of FN concentrations (supplementary material Fig. S3). Thus kindlin-2 targeting to adhesions is required for complete rescue of cell spreading, but the β1 extracellular domain also makes important contributions, presumably through enhanced avidity for ligand.

Role of kindlin-2 recruitment in integrin-β6-mediated cell spreading and focal complex formation

Our previous data suggest that localization of kindlin-2 to focal adhesions rescues the β1-KO phenotype in cells. This led us to ask whether a β6 chimera that can bind kindlin-2 would be able to rescue the β1-KO phenotype. To address this question, we generated the β6-β1 construct (extracellular and transmembrane domain of β6 with cytoplasmic domain of β1). β1-KO cells were infected with the β6-β1 rescue construct and the transduced cells were sorted by FACS to identify cell populations that expressed high levels of the chimeric rescue construct as judged by IRES–GFP expression. These cells were plated on fibronectin-coated coverslips and analyzed for the expression of β6 and vinculin and β6 and kindlin-2. The WT cells were well spread, displayed focal adhesions that had robust staining for β1 and vinculin (Fig. 8B) and β1 and kindlin-2 (Fig. 8C,C′). The KO cells did not express β1 (Fig. 8E), were not well spread, and kindin-2 did not target to the large β6-rich peripheral focal adhesions (Fig. 8F,F′). Interestingly, β6-β1-rescued cells were much better spread than the KO cells (Fig. 8J) and showed robust kindlin-2 targeting to focal adhesions (Fig. 8I,I′). Analysis of focal adhesion area revealed that the focal adhesions in β6-β1 cells were significantly smaller (1.4±0.2 μm²; n=1075) than those of KO cells (3.5±0.5 μm²; n=1051) (Fig. 8H) and comparable to that seen in WT cells 1.3±0.2 μm² (n=1005). Despite rescue of focal adhesion size and an increase in cell spreading in β6-β1 cells, cell area was not completely rescued to the WT level (Fig. 8J). This is consistent with our data showing that the β1 ectodomains contribute to the spreading process presumably because of the lower avidity of β6 for fibronectin compared with β1, as discussed in the previous section. Taken together, our data suggest that the inability of β6 to efficiently recruit kindin-2, coupled with its lower avidity for fibronectin, result in the reduced cell spreading seen in the β1-KO cells.

Fig. 8.

Recruitment of Kindlin-2 by β6 integrin mediates cell spreading and focal complex formation. (A) Surface expression of endogenous β1 integrin on WT cells assessed by FACS. (B,C) WT cells stained for β1, vinculin and kindlin-2 (C′) as indicated. (D) Surface expression of endogenous β1 integrin on KO cells assessed by FACS. (E,F) KO cells stained with β1, vinculin, β6 and kindlin-2 (F′) as indicated. (G) Cells expressing high levels of β6-β1 were sorted based on high expression of GFP. Cells rescued with β6-β1 (H,I) were stained with β6 vinculin and kindlin-2 (I′) as indicated. (J) Box plot of the distribution of the average cell area along the median (n=100) of WT, KO and β6-β1-rescued cells (***P<0.001, compared with KO cells). (K) Table showing the mean focal adhesions and cell surface areas of the WT, KO and β1-β6-rescued cells. Scale bar: 10 μm.

Discussion

Each of the mammalian kindlins have been shown to play important roles in integrin activation and signaling (Karaköse et al., 2010; Lai-Cheong et al., 2010; Larjava et al., 2008; Malinin et al., 2010; Meves et al., 2009; Plow et al., 2009). Keratinocytes express both kindlin-1 and kindlin-2. These proteins are highly conserved, exhibit overlapping functions and can, at least in part, functionally compensation for each other (He et al., 2011a). Here, using β1-integrin-deficient keratinocytes that adhere to fibronectin through αvβ6 integrins, we have uncovered functional differences between kindlin-1 and kindlin-2, demonstrated kindlin-1 binding to β6 integrins, revealed specificity in integrin β-tail binding of kindlins, and shown the importance of kindlin-2 targeting to focal adhesions for cell spreading. Furthermore, our data confirm that, despite extensive conservation between different integrin subunits and some partial redundancy in function, β1 integrins play unique, essential roles in matrix organization and intracellular signaling, and suggest that the ability of β1 integrins to bind kindlin-2 and recruit it to sites of extracellular matrix (ECM) adhesion is important for normal keratinocyte function.

Correct assembly of the ECM and signaling from the ECM are central to the normal function of the skin, and inappropriate expression of ECM receptors of the β1 integrin family affects multiple processes, including hair follicle morphogenesis, wound healing and cancer metastasis (Brakebusch and Fässler, 2005; Brakebusch et al., 2000; Legate et al., 2009; Raghavan et al., 2000; Wickström and Fassler, 2009; Wickström et al., 2011). Despite a compensatory upregulation of αvβ6 integrins, conditional ablation of β1 integrins in skin results in severe blistering and basement membrane defects (Brakebusch et al., 2000; Raghavan et al., 2000). In culture, β1-KO keratinocytes adhere to fibronectin through the upregulated αvβ6 integrins, but are defective in cell spreading, directed migration, and signaling and form large, peripheral, β6-rich focal adhesions (Raghavan et al., 2003). The αvβ6 integrins have a more restricted repertoire of ECM ligands than the β1 class integrins, but bind fibronectin well, suggesting that the defects in adhesion assembly, cell spreading and signaling of β1-KO cells plated on fibronectin reflects differential recruitment of intracellular signaling and adaptor proteins to β6 integrins. To address this question, we surveyed the localization of known focal adhesion proteins in wild-type and β1-KO keratinocytes. Vinculin, paxillin, FAK, talin and kindlin-1 were effectively targeted to focal adhesions in KO cells, but we noted a reduction in ILK targeting and defective adhesion targeting of kindlin-2. Similar defects in localization were seen in mouse skin, suggesting that one reason that β6 cannot substitute for β1 is its inability to recruit kindlin-2 to adhesions.

Kindlins are an evolutionary conserved family of adaptor proteins with key roles in integrin-mediated adhesion and signaling (Larjava et al., 2008; Meves et al., 2009; Moser et al., 2009). Kindlins have recently been shown to bind the cytoplasmic tails of β1, β2 and β3 integrins and the binding site was localized to the membrane-distal NPxY motif and a preceding threonine-containing motif in the integrin β-tail (Harburger et al., 2009; Ma et al., 2008; Moser et al., 2008). Here we extend these data to show that kindlin-1 also binds β6-tails. However, consistent with its defective recruitment to β6-mediated adhesions, we also observed that kindlin-2 bound very poorly, if at all, to β6-tails. To our knowledge, this is the first example of differential binding and recruitment of kindlin-1 and kindlin-2 to two different integrin subunits. Kindlin-1 and kindlin-2 exhibit 62% amino acid identity, and so their sharp difference in β6 binding is unexpected. However, chimeric analysis reveals that the region of β6 that mediates differential binding lies within or just after the second NPxY sequence, which fits our current understanding of the kindlin-binding site of integrin β-tails. High-resolution structural data on kindling–integrin interactions will be required to fully explain the basis for the differential β1 and β6 binding, but it is noteworthy that we previously revealed subtle differences between the interactions of kindlin-1 and kindlin-2 with β1 integrins (Harburger et al., 2009). Specifically, substitution of Ile651 with alanine strongly impairs kindlin-1 binding to integrin β-tails but the corresponding mutation in kindlin-2 (I654A) had no noticeable effect. Furthermore, deletion of the three C-terminal amino acids from β1 completely abrogated kindlin-1 binding, but allowed some residual kindlin-2 binding (~30%). Thus there is precedence for differences between kindlin-1 and kindlin-2 interactions with integrins, but the effect on integrin β6 binding is much more dramatic.

Both kindlin-1 and kindlin-2 are expressed in the skin. The phenotype of kindlin-1-knockout mice and Kindlin syndrome, caused by loss of kindlin-1, confirms an important role for kindlin-1 in skin and shows that even in the presence of endogenous levels of kindlin-2, kindlin-1 is required (Siegel et al., 2003; Ussar et al., 2008). Our data suggest that kindlin-2 also plays specific roles in keratinocyte function that are not fully compensated by endogenous kindlin-1. Specifically, our results indicate that expression and focal adhesion targeting of kindlin-2 is required for efficient cell spreading. This conclusion is based on the observation that similar to knockout of β1 integrin, kindlin-2 knockdown produces a spreading defect; furthermore, in β1-KO cells, re-expression of β1- or kindlin-2-binding chimeric integrins, but not non-binding chimeras, fully rescues cell spreading and restores focal adhesions to their normal size. In addition, β6-β1 chimeras capable of binding both kindlin-1 and kindlin-2 also rescue focal adhesion size and partially rescue cell spreading in β1-KO cells. It seems likely that the lack of a complete rescue relates to differences in integrin ectodomain affinity for fibronectin. Indeed the cell adhesion experiments that we performed with WT and KO cells on fibronectin indicate that the KO cells are approximately 30% less efficient at binding to fibronectin compared with the WT cells. These data are consistent with our previous findings that the KO cells adhere less well to fibronectin compared with the WT cells (Raghavan et al., 2003). Based on these findings, we suggest that the extracellular domain of β1 also contributes to cell spreading, presumably through increased avidity.

It is unclear why kindlin-1, which targets to adhesions in both β1-KO cells and in kindlin-2-knockdown cells, cannot fully compensate for the lack of kindlin-2 targeting, but it is possible that other kindling-specific binding partners exist that account for the differential effects. Our conclusion that kindlin-1 and kindlin-2 have independent functions in primary keratinocytes is fully compatible with published reports (He et al., 2011a) that the two kindlin genes also have overlapping roles in keratinocytes. Indeed, the similarities in sequence, known binding partners and subcellular localization, combined with the compelling data that cells lacking both kindlin-1 and kindlin-2 have a stronger phenotype than those lacking only a single kindlin (He et al., 2011a), make it clear that kindlins have considerable overlap in function. Nonetheless, each kindlin also has unique functions and, given the early embryonic lethality associated with knockout of kindlin-2 (Montanez et al., 2008), our results highlight the importance of generating and analyzing a conditional knockout of kindlin-2 in the skin to determine the in vivo significance of keratinocyte kindlin-2.

In summary, the availability of the β1-KO cells that adhere to the fibronectin matrix through αvβ6 integrins has allowed us to uncover important functional differences between kindlin-1 and kindlin-2 in keratinocytes. Furthermore, our data suggest that despite extensive conservation between different integrin subunits and some partial redundancy in function, β1 integrins play unique essential roles in the assembly and organization of the extracellular matrix and in intracellular signaling.

Materials and Methods

Cell culture

WT and β1-KO keratinocytes were grown in high calcium E medium (DMEM/>F12) in a 3:1 ratio with 15% FBS supplemented with insulin, transferrin, hydrocortisone, cholera toxin, triiodothyronone and penicillin-streptomycin at 32°C. The CHOA5 cells were grown in DMEM with 10% FBS and penicillin-streptomycin at 37°C.

GST pull-down assays

The β1, β6, mutant and chimeric constructs were expressed as GST fusion proteins in BL21 cells and purified using glutathione agarose beads (Amersham) according to the manufacturer's instructions. CHOA5 cells were transiently transfected, with the FLAG-tagged kindlin-1 and kindlin-2 plasmids (Harburger et al., 2009) using Lipofectamine 2000 (Invitrogen). Cell lysates from transfected CHOA5 cells or wild-type keratinocytes were resuspended in the lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, and 1% Triton X-100) containing protease and phosphatase inhibitors (Sigma). GST proteins were incubated with 100 μg of the cell lysates overnight at 4°C. Beads were washed in PBS, boiled and loaded onto SDS gels, transferred onto nitrocellulose and processed for western blotting using antibodies against FLAG (Sigma), kindlin-1 or kindlin-2. The western blots were scanned using the LICOR Odyssey Infrared Scanner.

Lentiviral infection of β1-knockout keratinocytes

All the β1 rescue constructs were cloned into the pLVV-IRES Zs Green lentiviral vector (Clontech). The lentiviruses were generated through transient co-transfection of the vector plasmid, the packaging construct pCMVD8.9 and the envelope coding plasmid pCMV-VSVG (from Soosan Ghazizadeh, SUNY, Stony Brook, NY) into 293T cells using Lipofectamine 2000 (Invitrogen). Supernatant containing the virus was collected after 72 hours and filtered through a 0.22 μm PES (Millipore) filter. The MOI was calculated by infecting 293T cells and ranged from 5×10⁵–1×10⁶ IU. β1-knockout keratinocytes were plated on the previous day at a low density (15–20% confluent). The medium was removed from the β1-KO keratinocytes and the filtered viral supernatant containing 8 μg/ml polybrene was added and incubated for 5 hours at 32°C. After 5 hours, the virus was removed and fresh E medium added.

FACS sorting of rescued cells

Infected cells (70–80% confluent) were harvested and subjected to FACS analysis. The infected cells were resuspended in a blocking solution (1% NGS in PBS) at a concentration of 10⁷ cells per ml. After blocking the cells for 30 minutes, the cells were stained with a phycoerythroin (PE)-conjugated β1 antibody (CD29, EBioscience). The cells were washed in PBS, resuspended in 1% NGS in PBS at a concentration of 10⁷ cells per ml and filtered. The cells were sorted using a Becton Dickinson FACS Aria sorter at low pressure (sheath pressure of 13 PSI) to avoid damaging the cells. Sorted cells were spun down at 1000 r.p.m. for 5 minutes at room temperature, resuspended in fresh E medium and plated.

Kindlin-2 knockdown

For the knockdown experiment, we purchased the TRC library based lentiviral shRNA clone for kindlin-2 and a scrambled probe from Sigma (St Louis, MO) and the lentiviruses were generated as described above and used to infect WT keratinocytes. The infected cells were grown and selected in the presence of 4 μg/ml puromycin and surviving. The cells that survived the selection were used for western blot analysis of kindlin-2 expression and in spreading assays.

Antibody staining

Approximately 5000–10,000 WT, KO, rescued and kindlin-2 knockdown cells were plated on glass coverslips coated with 10 μg/ml fibronectin for 23 hours. Cells were fixed with 4% paraformaldehyde for 10 minutes at room temperature (RT), permeabilized with PBST (PBS+ 0.2% Triton X-100) for 10 minutes and blocked with 5% normal donkey serum in PBST for 1 hour at room temperature. The cells were incubated with the primary antibody for 1 hour, washed five times with PBS and incubated with secondary antibodies coupled to Fluorescein isothiocyanate (FITC) and Rhodamine Red-X (RRX) (1:100; Jackson ImmunoResearch Labs, West Grove, PA). The nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI). For the skin sections, the primary antibody was incubated overnight at 4°C. The following antibodies were used: β1, talin, ILK, α6, FAK (all 1:100, Millipore), β6 (1:500, Biogen Idec/Stromedix, Boston), kindlin-1 (1:200, from Mary Beckerle, University of Utah, Salt Lake City, UT), kindlin-2 (1:500, from Cary Wu, University of Pittsburgh, Pittsburgh, PA), Migfilin (1:25, Proteintech), vinculin (1:100, Sigma), laminin 5 (1:200 from Bob Berguson, Harvard University, Cambridge, MA), keratin 14 (from Sat Sinha, SUNY, Buffalo, NY). Cells were visualized and photographed using an Olympus BX53 fluorescence microscope equipped with an UpanLFN 100× 1.3 NA oil lens. The Olympus DS2 software was used to acquire the images, which were then saved as TIF files and imported into Photoshop. The images were modified for brightness and contrast.

Cell adhesion assay on fibronectin

An eight-well glass chamber slide (Lab-Tek, Nunc) was coated with 0 μg/ml, 2 μg/ml, 6 μg/ml or 10 μg/ml (diluted in PBS) overnight. Wells were blocked with 10 mg/ml heat-inactivated BSA for 1 hour at 37°C. WT and KO cells (5×10⁴) were plated in quadruplicate and allowed to attach for 1 hour at 37°C. After removing the non-adherent cells, the cells were fixed in 4% paraformaldehyde, stained with DAPI and counted using a Nikon PFS-TE-2000 automated microscope. The experiment was performed in triplicate.

Surface area and focal adhesion area calculations

For each experiment, ~100 cells were photographed. The outline of each cell and focal adhesion (stained with vinculin) was marked using the Olympus DS software. The Average area of 100 cells and ~1000 focal adhesions was calculated and then used to generate the box plots. Statistical analysis was performed using ANOVA and Tukey–Kramer post-hoc test for all pair-wise comparisons (Prism 4 for Macintosh).