Integrin-linked kinase at a glance

In vivo studies have revealed that ILK is a ubiquitously expressed protein, whose predominant function is to organise the actin cytoskeleton during invertebrate and vertebrate development and homeostasis. In Caenorhabditis elegans, the ILK orthologue PAT-4 localises to integrins at muscle attachment sites. Deletion of the pat-4 gene causes a ‘paralysed at two-fold-stage’ (PAT) phenotype that is characterised by muscle detachment through defective integrin–actin linkage and early lethality (Mackinnon et al., 2002). In Drosophila melanogaster, a germline deletion of ILK leads to muscle detachment and lethality (Zervas et al., 2001). Mice lacking ILK die during the peri-implantation stage owing to a failure to organise the F-actin cytoskeleton in epiblast cells (Sakai et al., 2003). In addition to the constitutive deletion, the mouse ILK-encoding gene has been deleted in several organs and cell types using the Cre/loxP recombination system. The outcome of these studies has been extensively reviewed elsewhere (Rooney and Streuli, 2011; Ho and Bendeck, 2009; Hannigan et al., 2007; Wickström et al., 2010b).

Structurally, ILK has three different domains: five ankyrin repeats at the N-terminus, followed by a pleckstrin homology (PH)-like domain and a kinase-like domain at the C-terminus (Chiswell et al., 2008; Yang et al., 2009) (see poster). Although ILK was shown to directly interact with integrin cytoplasmic tails, it appears that the recruitment of ILK to integrins depends, at least in some cells, on kindlin-2 (also known as Fermt2 and Plekhc1) (Montanez et al., 2008; Chen et al., 2008), α-parvin (Fukuda et al., 2009) or paxillin (Nikolopoulos and Turner, 2001). Structural studies of ILK have revealed, however, that the proposed paxillin-interacting residues are buried within a polypeptide fold and thus are not directly accessible (Fukuda et al., 2009), suggesting that these residues indirectly contribute to paxillin binding. Before the recruitment of ILK to FAs, ILK forms a ternary complex with the two adaptor proteins Pinch and parvin (termed the IPP complex). Although it is not understood how the IPP complex forms, its formation ensures the stability of the individual components and their faithful targeting to the adhesion site (Zhang et al., 2002; Fukuda et al., 2003a). Mammals have two Pinch genes (Pinch-1 and Pinch-2; also known as LIMS1 and LIMS2, respectively), which encode proteins consisting of five cysteine-rich, zinc-binding LIM domains followed by a nuclear export signal. The first LIM domain of Pinch-1 and -2 binds to a concave surface that extends from the second to the fifth ankyrin repeat of ILK (Chiswell et al., 2008; Yang et al., 2009) (see poster). The three mammalian parvin isoforms (α-, β- and γ-parvin) are composed of an N-terminal polypeptide followed by two calponin homology (CH) domains, the second of which binds to the kinase-like domain of ILK (Tu et al., 2001; Fukuda et al., 2009) (see poster). As ILK can only bind one Pinch and one parvin isoform at the same time (Chiswell et al., 2008; Montanez et al., 2009), ILK is capable of being part of several distinct IPP complexes, each resulting in different signalling outputs (see poster and below).

The parvin proteins can interact directly with F-actin (Legate et al., 2006) or they can recruit actin-binding proteins, such as α-actinin – shown for β-parvin (Yamaji et al., 2004) – or vinculin, an interaction mediated through paxillin (Turner, 2000), which has been shown for α- and γ-parvin (Yoshimi et al., 2006). In addition, they control actin regulatory proteins such as testicular protein kinase 1 (TESK1), which can bind α-parvin and promote F-actin polymerisation through phosphorylation of cofilin (LaLonde et al., 2005). By contrast, β-parvin regulates actin dynamics through PAK-interactive exchange factor alpha (α-PIX, encoded by ARHGEF6), a guanidine exchange factor (GEF) for Rac1 and Cdc42 (Mishima et al., 2004). Finally, α-parvin has been shown to inhibit G-proteins by recruiting Cdc42 GTPase-activating protein (CdGAP, also known as RHG31 and Kiaa1204, and encoded by ARHGAP31) to FAs (LaLonde et al., 2006), and to negatively regulate Rho-associated protein kinase (ROCK)-driven contractility in vascular smooth muscle cells (Montanez et al., 2009).

Pinch-1 binds the Ras suppressor protein 1 (RSU1), which is important for integrin-mediated cell adhesion and spreading (Kadrmas et al., 2004; Ito et al., 2010). RSU1 is a negative regulator of growth-factor-induced Jun N-terminal kinase 1 (JNK1, also known as MAPK8) (Kadrmas et al., 2004). Taken together these findings suggest that the assembly of distinct IPP complexes in a given cell, together with the differential expression patterns of Pinch and parvin isoforms, provides a means for multiple alternative signalling outputs (see poster).

The most prominent subcellular localisation of ILK is in integrin adhesion sites. In recent years it has been reported that ILK is also present in additional subcellular regions and compartments where it might exert integrin-independent functions.

Keratinocytes, and probably other cells, employ ILK to capture microtubule (MT) tips to connect them to the cortical actin network (see poster). ILK-mediated MT capture occurs exclusively in nascent FAs and is mediated by recruitment of the large scaffold protein IQ-motif-containing GTPase-activating protein 1 (IQGAP1) (Wickström et al., 2010a). The capture of MT tips can be achieved either directly through binding of IQGAP1 to the MT tip protein CLIP170 (cytoplasmic linker protein 170; also known as CLIP1), or indirectly through IQGAP1-mediated recruitment of mammalian diaphanous homolog 1 (mDia1; also known as DRF1), which is also able to stabilise MTs. As both IQGAP1 and mDia1 are also able to bind F-actin, the ILK–IQGAP1–mDia1 complex connects MTs with actin tracks at β1-integrin-containing nascent adhesion sites (Wickström et al., 2010a) (see poster). Exocytotic carriers that are transported on MT tracks require a switch from MT-based to actin-based motility at the plasma membrane to pass through the cortical F-actin network and finally fuse with the plasma membrane. Thus, the connection of both networks by the ILK–IQGAP1–mDia1 complex at nascent adhesion sites is essential for the exocytosis of caveolar carriers (Wickström et al., 2010a). Consequently, ILK not only contributes to epithelial cell polarisation through actin remodelling, but also through vesicular trafficking and MT organisation.

Despite its prominent localisation in different integrin adhesion sites, ILK has also been observed in the nucleus of several cell lines, including COS-1 cells (Chun et al., 2005), MCF-7 cells (Acconcia et al., 2007), HeLa cells and keratinocytes (Nakrieko et al., 2008a) (see poster). The nuclear function of ILK, however, is still not well understood. In keratinocytes, nuclear ILK has been shown to induce DNA synthesis (Nakrieko et al., 2008a) and, in MCF-7 cells, it has been found to control the expression of the connector enhancer of kinase suppressor of Ras3 (CNKSR3) gene (Acconcia et al., 2007). It is known that CNKSR3 regulates the epithelial Na⁺ channel (ENaC) through inhibition of the MAPK kinase MEK1 (Ziera et al., 2009). However, the significance of ILK-regulated CNSKR3 expression is not understood.

It is also not well understood how ILK translocates into the nucleus. For example, it is not known whether the nuclear import of ILK depends on its N-terminus (Acconcia et al., 2007) or on a C-terminal nuclear localisation signal (Chun et al., 2005). The nuclear export of ILK requires the kinase-like domain (Acconcia et al., 2007; Nakrieko et al., 2008a) and is apparently controlled by the nuclear export factor CRM1, integrin-linked kinase-associated serine and threonine phosphatase 2C (ILKAP) and p21-activated kinase 1 (PAK1) (Acconcia et al., 2007; Nakrieko et al., 2008a).

ILK has been shown to serve as a scaffold for promoting the formation of cell–cell contacts (see poster) and the recruitment of tight junction proteins (Vespa et al., 2003; Vespa et al., 2005). Following treatment of cultured keratinocytes with Ca²⁺, they undergo differentiation. This process is accompanied by the translocation of ILK from FAs to cell–cell adhesion sites (Vespa et al., 2003). This translocation is known to require the N-terminal ankyrin repeats (Vespa et al., 2003); however, it is unclear whether Pinch-1 or -2 translocate together with ILK. In contrast to these in vitro findings, deletion of the Ilk gene in mouse keratinocytes neither affects cell–cell adhesion nor barrier function in the epidermis, but severely impairs keratinocyte migration on and adhesion to the epidermal-dermal basal membrane (BM), resulting in skin blistering, epidermal hyperthickening and hair loss (Lorenz et al., 2007; Nakrieko et al., 2008b).

A proteomic search for new ILK-interacting proteins identified a number of proteins, including several centrosome- and mitotic-spindle-associated proteins, such as α- and β-tubulin, the tubulin-binding proteins RUVBL1 and colonic and hepatic tumor overexpressed gene protein (ch-TOG, also known as CKAP5) (Dobreva et al., 2008) (see poster). Although ILK probably binds these proteins in an indirect manner (Fielding et al., 2008), it colocalises with them in centrosomes from interphase and mitotic cells where it has an essential role in controlling centrosome function during mitotic spindle organisation and centrosome clustering (Fielding et al., 2008; Fielding et al., 2011). The organisation of the mitotic spindle requires the kinase Aurora A and the association of ch-TOG with the centrosomal transforming acidic coiled-coil-containing protein 3 (TACC3), which in turn promotes the polymerisation and stabilisation of centrosomal MTs (Barr and Gergely, 2007). In ILK-depleted cells, Aurora A kinase, although active, is unable to phosphorylate and thus activate TACC3, resulting in disruption of mitotic spindles. Similarly, the clustering of supernumerary centrosomes in cancer cells is also achieved by the TACC3–ch-TOG complex in an ILK- and Aurora-A-dependent manner (Fielding et al., 2011). ILK associates with ch-TOG, but not with TACC3 or Aurora A. Therefore, it is not clear how ILK supports the phosphorylation of TACC3 by Aurora A. Similarly, it is also unclear how ILK is recruited to centrosomes. The centrosomal localisation of ILK requires RUVBL1 expression and occurs without the known ILK-binding partners, α-parvin and Pinch (Fielding et al., 2008). Finally, it is also not known why the treatment of cells with QLT-0267, a small chemical compound that binds to the ATP-binding site of ILK, is as effective as siRNA-mediated depletion of ILK in blocking the association of TACC3 with Aurora A (Fielding et al., 2008). The mechanistic interpretation of this work is based on the assumption that ILK acts as a kinase, which has been disproved by genetic and structural studies (see below). A potential explanation for the inhibitory effect of QLT-0267 could be an impairment of the stability of ILK (see the next section).

The experimental evidence that the kinase-like domain of ILK lacks catalytic activity is overwhelming (Wickström et al., 2010b). Although ILK was initially identified by Dedhar and colleagues as a serine/threonine kinase (Hannigan et al., 1996), it lacks several important motifs that are conserved in most kinases (Hanks et al., 1988) (see poster). Furthermore, genetic studies in flies, worms and mice have demonstrated that the putative kinase activity is not required for development and homeostasis (Zervas et al., 2001; Mackinnon et al., 2002; Lange et al., 2009). Despite this compelling evidence, many papers have been and are still published claiming that ILK is a bona fide kinase, with only marginal evidence at best.

The crystallisation of kinase-like domain of ILK in complex with the CH domain of α-parvin and its comparison to the kinase domain of protein kinase A (PKA) has provided a mechanistic explanation for why the kinase function of ILK is not executed (Fukuda et al., 2009). The catalytic activity of a kinase depends on a coordinated interplay of the N- and C-lobes and the catalytic loop of the kinase domain with the substrate and ATP. The N- and C-lobes and the catalytic loop of ILK show major differences to those of bona fide kinases that render the ‘kinase’ of ILK non functional: (1) The catalytic loop lacks important acidic and positively charged residues. The acidic residue (D166 in PKA), which polarises the hydroxyl group of the substrate and accepts its proton, is replaced in ILK with the uncharged alanine residue (A319) (Fukuda et al., 2009). (2) The positively charged residue in the catalytic loop (K168 in PKA), which stabilises the intermediate state of the phosphoryl transfer reaction by neutralizing the negative charge of the γ-ATP phosphoryl group, is replaced in ILK by N321 resulting in a misrouting of ATP to the C-lobe. (3) In the N-lobe, the ATP-binding p-loop captures ATP at a too great distance from the active centre (10Å), which precludes its movement towards the catalytic loop, and the lysine residue K220 contacts the α- and γ-ATP phosphoryl groups instead of the α- and β-phosphoryl groups resulting in an aberrant ATP orientation. (4) Furthermore, the C-lobe of ILK chelates ATP with only one instead of the expected two metal ions. The metal ion is bound by the aspartate residue (D339) of the DVK motif of ILK (a DFG motif in PKA), whereas the second potential metal-binding residue (S324) remains unoccupied. Another divergence from bona fide kinases is the coordination of the γ-ATP phosphoryl-group by the lysine residue (K341) of the DVK motif of the N-lobe, which usually is mediated by the catalytic loop (Fukuda et al., 2009). Nevertheless, despite this structural evidence, dissenting views are still expressed and the controversy rages on (Hannigan et al., 2011).

Thermodynamic and structural analysis of ILK mutants has revealed that the K220A and K220M mutations, previously described as affecting kinase function, destabilise the global ILK structure (Fukuda et al., 2011), thus reducing ILK stability and the binding of interaction partners such as α-parvin. This observation provides an explanation for the severe kidney defects observed in mice which either lack α-parvin expression or carry K220A or K220M mutations in ILK (Lange et al., 2009).

ILK is overexpressed in many types of cancer, and it has been reported that its depletion or inhibition with the small molecule inhibitor QLT-0267 inhibits anchorage-independent growth, cell cycle progression and invasion (Hannigan et al., 2005). The oncogenic effects of ILK have been attributed for the most part to the catalytic activity of the kinase domain resulting in the activation of protein kinase B (PKB, also known as Akt) and glycogen synthase kinase-3 beta (GSK3β), which in turn regulates the stability of proto-oncogenic β-catenin (Hannigan et al., 2005). The recent findings showing that mammalian ILK lacks catalytic activity and serves as a scaffold protein in FAs of mammalian cells (Lange et al., 2009; Fukuda et al., 2009) raise the question of how ILK mediates its oncogenic potential despite this functional twist.

One possibility is that ILK controls the activity of oncogenes, such as PKB by controlling their subcellular localisation. For example, the ILK-binding partners α- and β-parvin induce the recruitment of PKB to the plasma membrane where PKB mediates its oncogenic activity (Fukuda et al., 2003b; Kimura et al., 2010). An alternative possibility is that ILK and ILK-interacting protein(s) regulate oncogenic kinases by controlling the activity of phosphatases. This has been shown for Pinch-1, which binds to and inhibits protein phosphatase 1α (PP1α) resulting in sustained PKB phosphorylation and activity (Eke et al., 2010). Consequently, reducing the amounts of the IPP complex in FAs will concomitantly result in an increased PP1α activity and decreased PKB function. It is also conceivable that ILK exerts its oncogenic function through its ability to cluster supernumerary centrosomes in cancer cells, thereby preventing their genomic instability and death (Fielding et al., 2011). Finally, ILK might also promote oncogenesis by regulating gene expression in the nucleus or by modulating the assembly of ECM proteins, as shown for fibronectin (Wu et al., 1998), which has been reported to affect cancer development and invasion (Akiyama et al., 1995).

ILK research has been significantly advanced in the past years by the resolution the long-lasting debate regarding the catalytic activity of ILK and by identifying novel functions for ILK, many of which occur outside of FAs. However, most of the emerging functions of ILK (e.g. in the nucleus, at cell–cell adhesion sites and in the centrosome) have only been studied in cultured cells thus far and still await confirmation in vivo.

In addition, several basic functions of ILK in FAs are still unresolved, including the mechanism(s) of the recruitment of ILK to FAs, the role of ILK in FA maturation and as a potential stretch sensor (Bendig et al., 2006), and the turnover and modifications of ILK, to name a few. Whether the recruitment of ILK to FAs occurs through a direct association with the integrin cytoplasmic domains or indirectly, e.g. through binding to kindlins (Montanez et al., 2008) or paxillin (Nikolopoulos and Turner, 2001) is currently unclear. In this regard, it is also not known whether ILK binds or associates with all β-integrin tails or whether this association is more selective. A co-crystallisation of the kinase-like domain of ILK with β-integrin tails should help to answer some of these questions. Similarly, a structural analysis of the predicted PH domain of ILK would clarify whether it adopts a classical PH fold, as predicted in the original publication (Hannigan et al., 1996), or a different motif, whose function would then have to be determined. Zebrafish studies point to a stretch-sensing function for ILK in cardiomyocytes (Bendig et al., 2006). This observation raises the question of whether mechanical stress sensing by ILK is restricted to cardiomyocytes or whether it also occurs in other cells, and of how ILK is executing this function at the molecular level. Finally, it will be important to re-evaluate the role of ILK in cancer. Ideally, these experiments should be performed in an unbiased manner with tumour models in mice (e.g. colon cancer and mammary cancer models) that lack ILK expression, and are complemented by sophisticated in vitro studies with cells derived from the tumours.

It is obvious that despite the rapid progress in ILK research, many questions are still unanswered. Recent advances in imaging and proteomics combined with genetics, cell biology and biochemistry will make the years to come exciting for all ILK aficionados.

We thank Kyle Legate and Roy Zent for careful reading of the manuscript and Max Iglesias for help with preparation of the art work.