Molecular dissection of the ILK-PINCH-parvin triad reveals a fundamental role for the ILK kinase domain in the late stages of focal-adhesion maturation

Cell adhesion receptors of the integrin family assume fundamental roles throughout development and in adult organisms by mediating interactions between cells and the extracellular matrix that control migration, survival, proliferation, differentiation and matrix assembly (Hynes, 2002; van der Flier and Sonnenberg, 2001). To accomplish these diverse functions, integrins assemble a host of ancillary proteins into specialized adhesive structures with distinct morphologies, subcellular localization and signaling potential (Geiger et al., 2001). Focal complexes (FXs), focal adhesions (FAs) and fibrillar adhesions (FBs) are among the most widely studied, with FXs being the smallest adhesions that emerge from highly dynamic nascent adhesions (Choi et al., 2008) at the edge of extending lamellipodia and guide cell spreading. Shortly after their formation, FXs can either dissolve or connect to actin stress fibers and enlarge to form FAs that provide firm anchorage to the substrate. FBs, which are sites of fibronectin (FN) fibrillogenesis characterized by their elongated shape and more central location, are formed by translocation of α5β1 integrin dimers out of FAs along nascent FN fibers on the cell surface (Pankov et al., 2000; Zamir et al., 2000).

Several reports have described the diversity of these adhesive structures in terms of their composition, life span, integrin density and turnover, and dependence on acto-myosin-generated tension (Ballestrem et al., 2001; Zaidel-Bar et al., 2003; Zamir et al., 1999). With regard to their molecular content, the formation of FXs, FAs and FBs is accompanied by the hierarchical recruitment of different integrins and integrin-associated proteins (Zaidel-Bar et al., 2003). Typically, paxillin is recruited early into FXs, followed by vinculin, whereas zyxin and tensin are recruited after the complete maturation of FAs. The actin-binding protein tensin is particularly abundant in FBs, whereas levels of phosphotyrosine (PY) on resident proteins are significantly lower in these matrix-forming adhesions (Zamir et al., 1999).

Unraveling the signals that regulate the dynamic conversion of FXs to FAs and FBs is of primary importance for understanding integrin function. Previous studies from our laboratory have identified integrin-linked kinase (ILK) as a key regulator of the stability and/or turnover of FAs and FBs in fibroblasts and endothelial cells (Boulter et al., 2006; Vouret-Craviari et al., 2004). ILK is a modular scaffolding protein comprising an N-terminal array of ankyrin repeats, a short pleckstrin-homology (PH) domain and a C-terminal Ser/Thr kinase-like domain (Delcommenne et al., 1998). By means of its N-terminal ankyrin repeat, ILK binds to the first of five LIM domains in the adaptor protein PINCH1 (Chiswell et al., 2008; Li et al., 1999). In addition to integrin β subunits, the kinase domain interacts directly with other components of integrin-based adhesion plaques, including paxillin and the most C-terminal of the two calponin-homology (CH) domains of the adaptor protein α-parvin (also known as actopaxin and CH-ILKBP) (Nikolopoulos and Turner, 2002).

ILK, PINCH1 and α-parvin form a ternary complex termed IPP (ILK-PINCH-parvin) that localizes to both FAs and FBs and is essential for several integrin-dependent functions (Legate et al., 2006; Wu, 2004). Formation of the IPP triad occurs before, and is a prerequisite for, recruitment of ILK, PINCH1 and α-parvin to integrin adhesions (Zhang et al., 2002), implying a functional interdependency of these proteins. ILK, PINCH1 and α-parvin proteins are also to a large extent interdependent for their stability as depletion of one leads to extensive proteasome-mediated degradation of the other two (Fukuda et al., 2003), thus rendering functional dissection of the complex difficult to achieve.

We show here that the IPP complex controls maturation of adhesions by stabilizing the integrin-actin connection and providing a permissive platform for tensin recruitment. Furthermore, using a novel gain-of-function strategy for molecular dissection of this complex, we identify the ILK C-terminus and α-parvin as essential partners in this process.

It has been shown in previous investigations that ILK^–/– and PINCH1^–/– fibroblasts exhibit fewer and smaller FA than parental ILK^fl/fl and PINCH1^fl/fl cells (Sakai et al., 2003; Stanchi et al., 2005). To explore the molecular basis of this phenotype, we analyzed the composition of ILK- and PINCH1-null adhesions by immunostaining. As adhesion maturation occurs through sequential recruitment of proteins (Zaidel-Bar et al., 2003), cells spreading on FN were stained with antibodies against the FN-binding α5β1 integrin and various `early' and `late' adhesion components.

In ILK^fl/fl and PINCH1^fl/fl cells, the peripheral FXs and FAs were rich in PY, paxillin and vinculin. Adhesions located more centrally contained lower levels of PY but were enriched in tensin1, consistent with the late recruitment of this protein during FA maturation (Fig. 1A; and F.S., unpublished results). Zyxin, described as a late FA component, was present in large FAs at the tips of thick stress fibers. In accordance with its role as a mechanosensitive protein (Yoshigi et al., 2005), zyxin could also be detected along actin microfilaments, as determined by phalloidin staining (D.G., unpublished results).

ILK-null cells were strikingly devoid of the tensin-rich central adhesions with low PY content. Instead, they formed a fringe of peripheral adhesions strongly staining for phosphotyrosine and containing the `early' FA components paxillin and vinculin. Zyxin was only faintly visible in adhesions of the ILK-null cells and never present along stress fibres (Fig. 1A). PINCH1<sup>–/–</sup> cells exhibited an identical phenotype (D.G., unpublished results). Western blot analysis revealed a downregulation of both tensin1 and zyxin proteins after ILK or PINCH1 deletion (Fig. 1B). Decreased protein expression was accompanied by a decrease in mRNA encoding tensin1 but not zyxin (Fig. 1C). Interestingly, tensin1 but not vinculin was detectable in ILK co-immunoprecipitates from ILK<sup>fl/fl</sup> lysates (Fig. 1D), suggesting that the IPP complex forms a tighter link with tensin1 than with components of `early' adhesions.

Tensin has been implicated previously in FB formation (Pankov et al., 2000). Indeed, the downregulation of tensin1 observed in ILK- and PINCH1-null fibroblasts was accompanied by a lack of FBs. Not surprisingly, in the absence of these matrix-forming adhesions, we observed a major defect in FN fibrillogenesis, although no impact on FN secretion could be detected (supplementary material Fig. S1). We then sought to determine whether exogenous tensin could restore FA maturation in ILK- and PINCH1-null cells. To do so, a retrovirus encoding chicken tensin dually tagged with a Flag epitope and monomeric red fluorescent protein (mRFP) was transduced in ILK^fl/fl and PINCH1^fl/fl cells. Single clones expressing the transgene were isolated (ILKT^fl/fl, PINCH1T^fl/fl) and retroviral-mediated Cre recombination was performed to obtain ILKT^–/– and PINCH1T^–/– cells. Western blotting confirmed the expression of Flag-RFP-tensin, as well as the absence of ILK and PINCH1 proteins in ILKT^–/– and PINCH1T^–/– clones (Fig. 2A). In agreement with previous reports, ILK or PINCH1 deletion was accompanied by a trans-downregulation of PINCH1 and ILK, respectively (Fukuda et al., 2003; Li et al., 2005; Stanchi et al., 2005).

In ILKT^fl/fl and PINCH1T^fl/fl cells, localization of Flag-RFP-tensin was identical to that of the endogenous protein (mostly in centrally positioned adhesions containing high levels of α5 integrin but low amounts of PY, morphologically similar to mature FAs or FBs) (Fig. 2B). By contrast, in ILKT^–/– cells and PINCH1T^–/– cells, Flag-RFP-tensin accumulated in perinuclear aggregated structures and it was only barely detectable in FAs. As described above, in the absence of ILK or PINCH1, α5 integrin remained in peripheral PY-rich adhesions without colocalizing with Flag-RFP-tensin (Fig. 2B, quantified in Fig. 2C), suggesting that tensin availability is not a limiting factor for adhesion maturation.

Collectively, these data suggest that ILK or PINCH1 deletion, although permissive for FX and `early' FA formation, abrogates their maturation to `late' FAs and FBs, as defined by topology and recruitment of tensin and zyxin.

Real-time analysis of cell shape changes provided important insights into the mechanisms underlying impaired FA maturation in the ILK- and PINCH1-null cells. Both ILK^–/– and PINCH1^–/– cells undergo cycles of rapid membrane retraction and frequent cell rounding. Thus, at any given time, between 20-30% of the cells in the population are round (Fig. 3B; supplementary material Movies 1-4), suggesting that the adhesion defect in ILK- and PINCH1-null cells might stem from reduced maintenance of cell-matrix contacts. As previous work has shown that microtubule (MT) disruption with nocodazole stabilizes FAs (Kaverina et al., 1999), we tested whether FA stabilization by disruption of the MT network in ILK- and PINCH1-null cells would inhibit cell retraction. In control ILK^fl/fl and PINCH1^fl/fl cells, nocodazole-induced MT collapse was accompanied by an increase in cytoskeletal tension, clearly visible at 45 minutes by a change in cell morphology, an increase in FAs and a thickening of actin stress fibers (Fig. 3A; and F.S., unpublished results). In ILK^–/– and PINCH1^–/– cells, rather than reinforcing FAs, nocodazole significantly perturbed adhesion and cells rounded up within 2-10 minutes (Fig. 3B; supplementary material Movies 1 and 2).

Maturation of FAs under normal conditions and upon MT disruption is associated with increased cytoskeletal tension (Kirchner et al., 2003). To determine whether this underlies the sudden collapse of ILK- and PINCH1-null Fas, we treated cells with the ROCK inhibitor Y27632 to relax actomyosin-mediated contractility. ROCK inhibition had little effect on the spreading of ILK^fl/fl and PINCH1^fl/fl cells, whereas relieving endogenous force in ILK^–/– and PINCH1^–/– cells halted their frequent contractions and substantially reduced the number of round cells in the populations (Fig. 3B; supplementary material Movies 3 and 4).

To monitor the effect of endogenous cytoskeletal tension on tensin localization, we generated clones of ILK^fl/fl and PINCH1^fl/fl cells dually transduced with retroviruses encoding Flag-RFP-tensin and β-actin-GFP (ILKTA^fl/fl and PINCH1TA^fl/fl). Subsequently, knockout clones were derived by Cre recombination (ILKTA^–/– and PINCH1TA^–/–). As long-term fluorescence exposure was toxic to cells, we induced a rapid increase in cell contractility with nocodazole and filmed for short periods of time. Flag-RFP-tensin remained in stabilized FAs following drug treatment in the parental ILKTA^fl/fl (Fig. 3C; supplementary material Movie 5) and PINCH1TA^fl/fl cells (F.S., unpublished results). By sharp contrast, addition of nocodazole to ILKTA^–/– and PINCH1TA^–/–cells led to an abrupt retraction of actin stress fibers and centripetal movement of Flag-RFP-tensin. This retraction coincided with the perinuclear accumulation of Flag-RFP-tensin described above, suggesting that, after its initial localization in Fas, tensin is released from these structures together with F-actin (Fig. 3C, supplementary material Movie 6; and F.S., unpublished results). This was not attributable to overexpression of β-actin-GFP as phalloidin staining of ILKT^–/– and PINCHT^–/– cells also showed tensin aggregates colocalizing with short, speckled endogenous F-actin structures (F.S., unpublished results). Staining of cells fixed immediately after time-lapse analysis revealed that contracted ILKTA^–/– and PINCH1TA^–/– cells left substantial amounts of α5 integrin attached to the substrate (Fig. 3D, arrowhead; and F.S., unpublished results). Thus, in the absence of ILK or PINCH1, cytoskeletal tension disrupts the integrin–F-actin linkage, suggesting that ILK and PINCH might serve to relay endogenous force to adhesions as they mature.

Gene ablation is a powerful strategy for dissecting signaling cascades and establishing the functional hierarchy among different proteins within a pathway, but in the case of the IPP complex this analysis is complicated by the interdependency of the component proteins for their stability and localization in integrin plaques (Fukuda et al., 2003; Li et al., 2005; Stanchi et al., 2005; Zhang et al., 2002). Accordingly, western blot analysis showed that deletion of ILK or PINCH1 leads not only to the respective downregulation of PINCH1 or ILK proteins but also of α-parvin (Fig. 4A). Re-expression of GFP-tagged ILK in PINCH1^–/– cells, or GFP-PINCH1 in ILK^–/– cells, failed to restore a spread phenotype and FA targeting of the proteins (F.S., unpublished results), consistent with the notion that an intact ternary IPP is required for FA residence (Zhang et al., 2002). Therefore, the similar phenotype observed in ILK- and PINCH1-null cells presumably reflects the functional ablation of the entire complex. Not surprisingly, silencing of expression of α-parvin in ILK^fl/fl fibroblasts by small hairpin RNA reduced the levels of ILK and PINCH1 proteins (Fig. 4B), impaired cell spreading and perturbed FA maturation, as determined by tensin1 and zyxin staining (D.G., unpublished results).

To investigate further the role of the IPP complex in adhesion maturation and tensin recruitment, we employed a novel gain-of-function approach that involved targeting the IPP complex to a β integrin subunit by chimeric fusion. GFP-tagged β3 integrin was chosen for two reasons. First, for this strategy, we needed to use an integrin subunit that is normally enriched in PY-rich FXs and `early' FAs, but not in PY-poor `mature' FAs or FBs containing tensin (Zaidel-Bar et al., 2003). Second, this GFP-fusion has previously been shown to be fully functional (Ballestrem et al., 2001) and in fibroblasts it forms a single integrin heterodimer (αvβ3) that can be monitored by fluorescence microscopy. Thus, a set of retroviral constructs encoding ILK, PINCH1 or α-parvin fused to the cytoplasmic tail of β3-GFP was constructed. The cellular distribution of unmodified β3 integrin was followed using retroviral constructs encoding the β3 subunit tagged with GFP or mRFP.

To validate this approach, NIH3T3 cells with normal ILK and PINCH1 levels were transduced with the β3-RFP retrovirus, then a single clone was isolated (3T3/β3-RFP) and transduced with vector encoding β3-GFP or one of the `β3-GFP-IPP chimeras' (β3-GFP-ILK, β3-GFP-PINCH, β3-GFP-parvin).

Western blot analysis using antibody against GFP showed bands of the expected sizes and similar intensities for each construct (Fig. 4C). Targeting of the entire IPP complex to each of the β3-GFP-IPP chimeras was verified by immunoprecipitation with antibody against GFP followed by western blotting to detect the presence of endogenous ILK, PINCH1 and α-parvin proteins. The endogenous proteins were undetectable in immunoprecipitates of integrin β3-GFP-expressing cells, whereas substantial amounts co-precipitated with their chimeric partners (Fig. 4D).

The distribution of the constructs was analyzed in cells plated on glass coverslips and stained with antibody against PY. In control cells, β3-GFP and β3-RFP colocalized completely, and substantial overlap was observed with PY staining, demonstrating that the fluorescent tags did not differentially affect localization of the integrin subunit (Fig. 4E). By contrast, addition of the IPP complex clearly altered its localization. Thus, the three β3-GFP-IPP chimeras were enriched in adhesions with weak PY staining that localized more centrally in the cells, or in what appeared to be retracting cell edges (Fig. 4E, arrowheads). Their accumulation in these older adhesions beneath the cell body and at the trailing edge of motile cells (Fig. 4F) suggests that they display a greater stability than adhesions formed by β3-RFP or β3-GFP. When expressed at slightly higher levels (as judged by fluorescence intensity), the β3-GFP-IPP chimeras, but not the β3-GFP control, completely replaced β3-RFP in all the adhesions (F.S., unpublished results).

We then assessed whether the re-deployment of β3-GFP following its fusion with IPP components coincided with the recruitment of tensin. To do this, β3-GFP and the β3-GFP-IPP chimeras were transduced in a clone of NIH3T3 cells expressing Flag-RFP-tensin. Flag-RFP-tensin colocalized poorly with β3-GFP/PY, whereas it was highly enriched in the PY-poor adhesions formed by the β3-GFP-IPP chimeras. Hence, fusion of PINCH1, ILK or α-parvin to the cytoplasmic tail of β3 integrin leads to its tight association with the entire IPP complex and prompts it to form tensin-rich PY-poor adhesions.

To determine whether the IPP-coupled integrins play a causal role in tensin recruitment and the maturation of late FAs, we expressed the β3-GFP-IPP chimeras, or the β3-GFP control, in PINCH1T^–/– cells (see Fig. 2) that lack these structures. Western blot analysis revealed similar expression levels for each construct and, in some cases, a rescue of endogenous ILK or α-parvin (Fig. 5A). Thus, in the PINCH1-null background, expression of β3-GFP-PINCH restored ILK and α-parvin to control levels, expression of β3-GFP-ILK rescued α-parvin, and β3-GFP-parvin partially rescued ILK. The localization of the β3-GFP-IPP chimeras in PINCH1T^–/– cells was analyzed in cells stained for several FA markers. Similar to the NIH3T3 cells described above, β3-GFP remained in peripheral PY-rich adhesions, without rescuing the defective recruitment of Flag-RFP-tensin (Fig. 5B). Conversely, expression of β3-GFP-PINCH prompted the abundant formation of small adhesions containing tensin distributed throughout the cell body. These structures could easily be distinguished from the perinuclear tensin aggregates observed in the PINCH1T^–/– cells as they contained αv integrin and low, but detectable, amounts of PY, vinculin (Fig. 5B,C,E) and paxillin (F.S., unpublished results). However, they were devoid of α5 and β1 integrins (Fig. 5D; and F.S., unpublished results). Surprisingly, in this PINCH1-null background, both β3-GFP-ILK and β3-GFP-parvin chimeras exhibited an identical distribution and exerted similar effects on the localization of tensin (Fig. 5E), suggesting that PINCH1 might not be directly required for this function. Treatment of cells with two different αv antagonists (Maubant et al., 2006) had no effect on the normal tensin-rich adhesions of PINCH1^fl/fl cells but dissolved the tensin-rich adhesions formed in PINCH1^–/– cells by the β3-GFP-IPP chimeras, suggesting that they function as bona fide αvβ3 adhesion receptors (supplementary material Fig. S2). Thus, constitutive targeting of the IPP complex to β3 integrin selectively confers upon αvβ3 heterodimers the ability to mature tensin-rich adhesions in an IPP-null background.

In an attempt to functionally dissect the IPP complex, we built a set of retroviral constructs encoding deletion mutants of the β3-GFP-IPP chimeras including two integrin β3-GFP-PINCH1 chimeras containing only the LIM1 domain (β3-GFP-LIM1) or lacking it (β3-GFP-ΔLIM1), two β3-GFP-ILK chimeras lacking the ankyrin repeat (β3-GFP-ΔAnk) or the kinase domain (β3-GFP-ΔKin) and two β3-GFP-α-parvin chimeras lacking the CH1 (β3-GFP-ΔCH1) or CH2 (β3-GFP-ΔCH2) domain. This new set of constructs was transduced in 3T3/β3-RFP cells, then protein expression and the ability of the various IPP domains to associate with endogenous members of the IPP complex were determined by western blotting and co-immunoprecipitation (Fig. 6A,B). Fig. 6D schematizes the domains of PINCH1, ILK and α-parvin that mediate IPP complex interactions and the results of the co-immunoprecipitation assays. As expected, β3-GFP-LIM1, but not β3-GFP-ΔLIM1, co-immunoprecipitated endogenous ILK and α-parvin proteins; β3-GFP-ΔAnk and β3-GFP-ΔKin co-precipitated only α-parvin and only PINCH1, respectively; β3-GFP-ΔCH1, but not β3-GFP-ΔCH2, co-immunoprecipitated both endogenous ILK and PINCH1.

The ability of the various IPP domains to drive adhesion maturation, as assessed by displacement of β3-GFP-containing adhesions with respect to control β3-RFP-tagged adhesions, was quantified by analyzing 16-bit grayscale-ratio images obtained by dividing GFP by RFP images of single cells. Identical distribution of GFP and RFP-tagged constructs yields a constant GFP:RFP ratio in all the FAs of a given cell (visualized as uniform gray intensity), whereas a different distribution leads to local variations of the ratio (and non-uniform gray intensity). Metamorph software was used to analyze the intensity standard deviation (i.s.d.) values on a set of ratio images of single cells from each population (supplementary material Fig. S3). Images from control cells expressing β3-GFP gave low i.s.d. values, in the range of the background noise of the detection system. Conversely, i.s.d. values were significantly higher in images from cells expressing the full-length β3-GFP-IPP chimeras, reflecting their differential distribution with respect to the β3-RFP internal control (Fig. 6C). In the case of the deletion mutants, high i.s.d. values were obtained for β3-GFP-LIM1, β3-GFP-ΔAnk and β3-GFP-ΔCH1 chimeras, whereas low i.s.d. values were observed for β3-GFP-ΔLIM1, β3-GFP-ΔKin and β3-GFP-ΔCH2 chimeras. The same pattern for tensin recruitment to the chimeric integrins was observed, using a clone of NIH3T3 cells expressing Flag-RFP-tensin (F.S., unpublished results).

Combined, these observations indicate that a defined region of the IPP complex, comprising the C-terminal domain of ILK and/or the CH2 domain of α-parvin, represents the minimal functional unit required for tensin recruitment and integrin redistribution. Either fused directly to β3-GFP or recruited through interaction with other fused IPP components, these domains are associated with all the integrin β3 chimeric constructs that exhibit a shift in localization, whereas they are absent in all of those that did not (Fig. 6D).

To extend these observations, β3-GFP-IPP deletion mutants were expressed in PINCHT^–/– and ILKT^–/– cells. First, we examined the integrity of the IPP complex (i.e. stability of the IPP components) in the different cells by western blotting. Results from these analyses are shown in the top of Fig. 7, and schematic representations of the IPP components present in each cell line and their resulting interactions are pictured below. Thus, expression of the ILK-binding β3-GFP-LIM1 PINCH1 chimera in PINCH1^–/– cells restored endogenous ILK levels, whereas β3-GFP-ΔLIM1 did not. The ILK-binding β3-GFP-ΔCH1 α-parvin chimera showed a similar effect, albeit more moderate, whereas the β3-GFP-ΔCH2 chimera did not. In both PINCH1- and ILK-null cells, the α-parvin-binding kinase domain of ILK (β3-GFP-ΔAnk chimera), but not the N-terminal ankyrin repeat region (β3-GFP-ΔKin chimera), rescued endogenous α-parvin levels, whereas expression of the PINCH1-binding ILK ankyrin repeats (β3-GFP-ΔKin chimera) rescued PINCH1 levels in ILK^–/– cells. Finally, β3-GFP-LIM1 caused an increase in endogenous α-parvin protein in PINCH1^–/– but not in ILK^–/– cells, indicating that its effect on α-parvin protein stability is due to the rescue of endogenous ILK (schematized in Fig. 7).

Together with our co-immunoprecipitation data above, these observations show that stability of the ILK, PINCH1 and α-parvin proteins is intimately linked to their direct interaction.

Next, the effect of the β3-GFP-IPP deletion mutants on tensin localization was examined in PINCH1T^–/– and ILKT^–/– cells. Consistent with our observations in NIH3T3 cells, only the deletion mutants β3-GFP-LIM1, β3-GFP-ΔAnk and β3-GFP-ΔCH1 retained the ability to form tensin-rich FAs in PINCH1T^–/– cells (supplementary material Fig. S4A).

In the ILKT^–/– cells, only the β3-GFP-ΔAnk chimera prompted the formation of tensin-rich adhesions, whereas all the other constructs, including β3-GFP-LIM1 and β3-GFP-ΔCH1, did not (supplementary material Fig. S4B). Similarly, the full-length β3-GFP-PINCH and β3-GFP-parvin chimeras did not rescue FA recruitment of tensin in ILKT^–/– cells (F.S., unpublished results and Fig. 8F), indicating that targeting the ILK kinase domain to integrins is strictly required for this event. Interestingly, ILKT^–/– cells expressing β3-GFP-ΔAnk showed increased spreading (Fig. 8A), which was not observed in ILKT^–/– cells expressing the other constructs.

Besides interacting with members of the parvin protein family, the ILK C-terminal domain interacts with paxillin (Nikolopoulos and Turner, 2002) and has been reported to possess Ser/Thr kinase activity (Hannigan et al., 1996). We next introduced point mutations in the β3-GFP-ΔAnk chimera that have previously been shown to impair these functions. The first mutation, V386G-T387G was shown to abrogate the paxillin-binding site (PBS) of ILK (Nikolopoulos and Turner, 2002) and the second, K220G, ablates a lysine residue in the ILK ATP-binding site reportedly essential for kinase activity (Yamaji et al., 2001).

Similar to the β3-GFP-ΔAnk construct, expression of the PBS mutant in ILKT^–/– cells increased cell spreading, albeit to a lesser extent than the wild-type construct, whereas the K220G mutant was not competent for this effect (Fig. 8A). Interestingly, both the effect on cell spreading and on the ability to prompt the formation of tensin-rich adhesions correlated with the capacity of the mutant constructs to bind to α-parvin, observed in co-immunoprecipitation experiments, suggesting that α-parvin is required for ILK-dependent FA maturation (Fig. 8B,C).

To test this, α-parvin was abrogated by small hairpin RNA knockdown in ILKT^–/– cells expressing the β3-GFP-ΔAnk construct (Fig. 8D). Following α-parvin depletion, β3-GFP-ΔAnk remained in peripheral adhesions and no longer rescued the defective recruitment of Flag-RFP-tensin (Fig. 8E). As mentioned above, the expression of β3-GFP-parvin did not rescue adhesion maturation (Fig. 8F). Combined, these observations point to a requirement for both ILK and α-parvin for this effect.

With respect to paxillin binding, no paxillin was detectable in the co-immunoprecipitates of either wild-type or mutant constructs, suggesting that the interaction of the ILK kinase domain with paxillin might be weaker than its interaction with α-parvin. Indeed, paxillin was detectable in the adhesions of all the cells, including the tensin-rich adhesions induced by expression of the PBS mutant, indicating that its localization does not depend upon direct binding to ILK (F.S., unpublished results).

Here, we sought to investigate the role of ILK, PINCH1 and α-parvin in regulation of integrin-based adhesions using unique cellular models and chimeric proteins designed to dissect their function. Previous studies have documented the presence of the early FA components paxillin and vinculin in adhesions of ILK- and PINCH1-knockout cells (Sakai et al., 2003; Stanchi et al., 2005). We extended this work to include analysis of `late' FA components, including tensin1, an F-actin capping and binding protein (Lo et al., 1994) and zyxin, a LIM-domain mechanosensitive protein (Yoshigi et al., 2005). Our results show that ablation of the IPP complex, by ILK or PINCH1 deletion, impacts on the recruitment of both tensin1 and zyxin to FAs, establishing these proteins as downstream targets of the IPP complex. Intriguingly, in ILK- and PINCH1-null cells, both tensin1 and zyxin proteins are downregulated by mechanisms that remain to be determined and that, in the case of tensin1, include a decrease in mRNA levels.

By virtue of its actin-binding activity, tensin has been proposed to be an important regulator of adhesion dynamics. Indeed, negative regulation by tensin of actin assembly by capping the barbed end of filaments (Chuang et al., 1995) could provide a mechanism to regulate the transmission of force between the actin cytoskeleton and FAs. However, restoring tensin expression alone in ILK- and PINCH1-deficient cells was not sufficient to drive FA maturation. As our study revealed, escalating force at the cytoskeletal membrane interface in ILK^–/– and PINCH1^–/– cells leads to a disruption of the integrin–F-actin linkage, even in presence of ectopically expressed tensin (Flag-RFP-tensin). In light of these findings, together with the established requirement for actomyosin-generated tension during FA formation or maintenance and integrin translocation (Pankov et al., 2000; Zamir et al., 1999; Zamir et al., 2000), we propose that the failure of IPP-null adhesions to sustain endogenous tension might explain their inability to support FA maturation and tensin recruitment. Colocalization between ILK and tensin might result from `activation' of the IPP complex by an unknown mechanism, possibly involving α-parvin and F-actin. This finding would support the `conformational switch' model of force-induced focal contact growth proposed by Geiger and Bershadski (Geiger and Bershadsky, 2001), in which force applied to specific adhesion-associated molecules might induce their conformational activation.

The novel strategy of imposed integrin localization of the IPP components used here allowed us: (1) to overcome the interdependency of the IPP proteins for their integrin association; (2) to assess the effects of IPP targeting to a specific integrin; and (3) to evaluate the functional contribution of individual IPP components to adhesion maturation. We found that, in contrast to β3-GFP alone, fusion of β3-GFP with ILK, PINCH1 or α-parvin confers these chimeras with the ability to reconstitute the entire IPP triad at their cytoplasmic tails. The resulting IPP-bound αvβ3 dimers form adhesions that resemble mature FAs (as defined by their morphology, localization and relative abundance of tensin), both in NIH3T3 fibroblasts, expressing the complete set of endogenous IPP components, and in ILK- and PINCH1-null cells. Systematic analysis of motifs within the IPP complex (β3-GFP-IPP deletion mutants) for their ability to drive the formation of these mature FA structures allowed us to identify the ILK kinase domain as a key requirement for this function.

An important finding using this strategy is that the IPP complex, through the ILK kinase domain, can selectively control the organization of adhesive complexes and the dynamics of the integrin to which it associates. Thus, although the β3-GFP-IPP chimeras can form tensin-rich FAs in PINCH1^–/– cells, their expression in this PINCH1-deficient background does not rescue the defective formation of these adhesions by α5β1 or by other β1 heterodimers that are unable to reconstitute the ternary complex. In light of our results here, and our previous findings that endogenous ILK colocalizes with the α5 integrin subunit (Vouret-Craviari et al., 2004), we propose that the process by which IPP binding to ligand-engaged α5β1 integrin prompts tensin recruitment and linkage to the actin cytoskeleton might underlie the α5β1 integrin segregation during FA maturation and FB formation. Concerning FB formation and FN assembly, the molecular mechanisms that regulate the binding of IPP to α5β1 are largely unknown. It has been shown using monoclonal antibodies that stepwise conformational changes in α5β1 integrin are associated with the process of FB formation and FN fibrillogenesis (Clark et al., 2005). We speculate that these changes, accompanied by ligand occupancy, might be linked to transitional events involving IPP association. It is noteworthy that, in our system, expression of the β3-GFP-IPP chimeras failed to drive complete FB maturation and FN fibrillogenesis (F.S., unpublished results) probably owing to intrinsic differences in FN binding activity of αvβ3 and α5β1 integrin, differences in their associated cytoplasmic protein complexes and/or co-receptors.

Current knowledge together with our present results suggests that recruitment of the IPP complex to integrins is subject to tight regulation that occurs at multiple levels. A first step involves the regulation of IPP complex formation. A second regulatory mechanism relies on proteasome-mediated degradation of the IPP components (Fukuda et al., 2003), which, as shown here, is intimately linked to the direct interaction between ILK, PINCH1 and α-parvin. Degradation of unbound IPP members could represent a `check-point' for control of dominant-negative effects of accumulating `incomplete' IPP complexes. Once a complete IPP complex has formed, a third level of regulation involves the establishment of parallel interactions between IPP components and other network proteins. The novel finding of our study that tensin and zyxin levels are sensitive to IPP components adds another layer of complexity to the tight control of adhesion dynamics.

The kinase domain of ILK was identified here as a key functional motif of the IPP complex; however, differences between signaling and structural or scaffolding functions of ILK are difficult to distinguish. Indeed, ILK kinase activity has been documented in numerous studies (reviewed by Legate et al., 2006). Nonetheless, the absence of the catalytic HRD and DFG motifs and the ability of a kinase-dead ILK to rescue ILK-null flies and worms have led to the classification of ILK as a pseudokinase with a scaffolding role (Boudeau et al., 2006; Legate et al., 2006). In our study, mutations in the ILK kinase domain of the β3-GFP-ΔAnk chimera ablating either paxillin (V386G-T387G) or ATP (K220G) binding impacted to varying degrees on its ability to rescue adhesion maturation. Mutations at these sites have previously been characterized in other systems, yet interpretation of their effects in terms of signaling has been complicated by their collateral impact on the FA targeting of ILK and, probably, of the whole IPP complex (Nikolopoulos and Turner, 2002). In the present study, this problem has been circumvented by chimeric fusion with β3-GFP to ensure FA localization.

Turning to the role of paxillin, its phosphorylation on two tyrosine residues (Y31 and Y118) was shown to control integrin adhesion dynamics by regulating recruitment of focal adhesion kinase (FAK) (Zaidel-Bar et al., 2007). As the kinase domain of ILK binds to the first leucine-rich domain (LD) of paxillin (Nikolopoulos and Turner, 2001) – that is, near Y31 – it could participate in this paxillin-dependent control of adhesion dynamics. To date, the consequence of paxillin phosphorylation on ILK binding is not known. However, our results herein that a partial rescue of the ILK^–/– phenotype occurred in cells expressing the PBS mutant (V386G-T387G) would argue that the observed effects of the ILK kinase domain do not depend exclusively upon direct interaction with paxillin. Indeed, although paxillin silencing in ILK^fl/fl cells abrogates the FA localization of FAK and fibronectin fibrillogenesis, this defect can be rescued with an ILK-binding paxillin mutant, further indicating that the direct interaction between ILK and paxillin is not essential (C.G., unpublished results). Alternatively, paxillin phosphorylation could regulate adhesion dynamics by controlling FAK and IPP recruitment in parallel. It is noteworthy that the CH2 domain of α-parvin can also bind to paxillin, and recent structural evidence indicates that paxillin LD1 might bind simultaneously to ILK and α-parvin, forming a three-way ternary complex (Wang et al., 2008).

With regards to the catalytic function of ILK, our findings that the ATP-binding-deficient mutant (K220G) did not co-precipitate α-parvin from cell lysates, together with the fact that α-parvin silencing abrogated the observed effects of the ILK C-terminus, indicate that recruitment of this actin-binding partner is essential for the adhesion maturation function of ILK.

Studies in invertebrates first suggested that ILK might behave as an adaptor that links integrins to the actin cytoskeleton (Mackinnon et al., 2002; Zervas et al., 2001). Our results suggest that, in vertebrates, ILK might control FA maturation in a similar way by creating a `docking platform' that connects F-actin through α-parvin with other actin-binding FA proteins, thereby reinforcing integrin-cytoskeleton linkage. Possible candidates for this cooperation include vinculin or filamin, which can be indirectly associated with ILK by their respective binding to paxillin or MIG2/kindlin-2 and migfilin (reviewed by Legate et al., 2006). As opposed to these early FA components, tensin is a relatively late addition to FAs with multiple actin-binding sites that has previously been shown to promote FB formation and FN assembly (Lo et al., 1994; Pankov et al., 2000). To date, our investigation is the first to establish a functional relationship between tensin and ILK in vertebrate cells, although similar findings in an invertebrate model suggest that this connection is conserved through evolution (Torgler et al., 2004). Similar to ILK- and PINCH1-null fibroblasts, both ILK and tensin Drosophila mutants display phenotypes that suggest a defect in the stabilization, but not in the initial formation of, integrin adhesions (Torgler et al., 2004; Zervas et al., 2001). Such conservation across evolution and complexity of regulation highlights the essential role of the IPP complex in integrin-dependent functions.

Retroviral expression constructs encoding integrin β3-GFP, integrin β3-RFP and the IPP fusions were derived from a β3-GFP construct (Ballestrem et al., 2001). Flag-RFP-tagged tensin was derived from a construct for transient expression of GFP-tagged chicken tensin (Zamir et al., 2000). Mutagenesis was performed using the QuikChange II site-directed mutagenesis kit (Stratagene). All constructs were subcloned in the retroviral expression vector pCLMFG (Naviaux et al., 1996).

All cells were cultured in DMEM supplemented with 10% fetal bovine serum for ILK mouse kidney fibroblasts (Sakai et al., 2003), PINCH1 mouse embryonic fibroblasts (Stanchi et al., 2005) and Phoenix ecotropic cells (ATCC SD-3444), and 10% bovine serum for NIH3T3 cells.

Phoenix ecotropic cells were transfected using the calcium phosphate method. Supernatants containing the retroviruses collected 48 and 60 hours post-transfection were supplemented with 5 μg/ml polybrene (Sigma) for infection of mouse fibroblasts.

Small hairpin lentiviral plasmids from Sigma (MISSION shRNA clones TRCN0000112665-69) were used for stable silencing of RNA encoding α-parvin. Viral particles were produced in 293 cells by transient co-transfection with the silencing vector (pLKO.1-puro backbone), the virion packaging system p8.91 (Didier Trono, EPFL Lausanne, Switzerland) and the envelope plasmid pMD2.G (Addgene plasmid 12259, Addgene, Cambridge, MA) generated in the laboratory of D. Trono.

Cre recombination was performed by infecting ILK^fl/fl and PINCH1^fl/fl fibroblasts with a self-excising Cre retrovirus (Silver and Livingston, 2001), prepared as described above. Cells were cloned and screened by PCR for gene deletion (Sakai et al., 2003; Stanchi et al., 2005).

Total RNA was isolated from cells using the Trizol reagent from Invitrogen. Semi-quantitative reverse-transcriptase polymerase chain reaction (RT-PCR) was performed using the OneStep RT-PCR kit from Qiagen. Amplification of 36B4 mRNA (GenBank accession number M17885) served as internal control. Forward and reverse primers used were, respectively, 5′-CGCCCCTGCTAAGCTCTTT-3′ with 5′-TCCATCCATACAGGGGATTC-3′ for tensin1 and 5′-CAGTGCTTCACCTGTGTGGT-3′ with 5′-CTGTGGTCTGGAAGGAGGTC-3′ for zyxin.

Reagents, unless specified, were from Sigma. Tissue-culture plasticware was from Nunc. The primary antibodies used have been described previously (Chu et al., 2006; Li et al., 2005; Stanchi et al., 2005; Vouret-Craviari et al., 2004), except anti-β-actin mAbcam 8226 (Abcam), anti-phosphotyrosine clone 4G10 (Upstate Biotechnology), anti-tensin1 (kindly provided by Su Hao Lo, University of California Davis, CA), anti-zyxin clone B-5-1-2 (Synaptic Systems, Göttingen, Germany), anti-α-tubulin clone 164D4 (Zymed), anti-integrin αv (Chemicon). Antibodies conjugated to Alexa488, Alexa546 and Alexa647 were from Molecular Probes, and HRP-conjugated antibodies were from Promega.

Immunoprecipitation, preparation of cell lysates, protein quantification and western blotting were performed as described previously (Stanchi et al., 2005), with the exception of anti-ILK immunoprecipitation, which was performed on cells lysed in NP40 lysis buffer (25 mM HEPES, 150 mM NaCl, 1 mM EGTA, 10 mM sodium pyrophosphate, 10 mM NaF, 5 mM sodium orthovanadate, 5 μg/ml leupeptin, 0.15 μM aprotinin, 1 mM AEBSF, complete, mini, EDTA-free; protease inhibitor cocktail tablet (Roche Diagnostics), 10% glycerol, 0.5% NP40) following brief sonication.

Cells were plated on glass coverslips uncoated or, when specified, coated with 10 μg/ml FN, fixed after 4 hours and stained as described previously (Vouret-Craviari et al., 2004).

Samples were observed through 63× (1.4 NA) oil or 10× (0.25 NA) air objectives on an Axiovert 200M inverted microscope (Carl Zeiss) equipped with an incubator chamber and a CoolSnap HQ cooled charge-coupled-device camera (Roper Scientific, Evry, France); all images were acquired and analyzed using MetaMorph Imaging System (Universal Imaging).