The extracellular matrix (ECM) provides the structural framework for the formation of tissues and organs. The ECM binds to substrate-adhesion molecules on the surface of cells and influences various intracellular signalling pathways that regulate survival, proliferation, polarity and differentiation. An important family of adhesion molecules that bind to the ECM are the integrins. Integrins are heterodimeric transmembrane molecules that consist of 
and 
subunits, and they are composed of large extracellular domains and relatively small cytoplasmic domains1. Integrins can switch between active and inactive conformations. In the inactive state, integrins have a low affinity for ligands. Intracellular signalling events such as protein-kinase-C stimulation can prime the integrins, which results in a conformational change that exposes the ligand-binding site (Fig. 1). Ligand binding activates signalling cascades that lead to the assembly of a multiprotein complex at the site of cell adhesion to the ECM. These events have two important impacts on the cell: they forge a connection between the ECM and the actin cytoskeleton, and they alter the fluxes of many intracellular signalling pathways.
Figure 1 | Biogenesis of focal adhesions.
a | Many integrins that are not bound to the extracellular matrix (ECM) are present on the cell surface in an inactive conformation, which is characterized by 'bent' extracellular domains that mask the ECM-binding pocket. This conformation is stabilized by interactions between integrin transmembrane domains, membrane-proximal extracellular domains and a salt bridge between the cytoplasmic domains. b | When talin is recruited to the plasma membrane and activated in association with phosphatidylinositol phosphate kinase type-I
(PIPKI), it binds to the cytoplasmic tail of integrins120. This interaction separates the cytoplasmic domains and induces the integrins to adopt the 'primed' conformation. c | The integrin extracellular domains extend and unmask the ligand-binding site, allowing the integrin to bind specific ECM molecules. The separated integrin cytoplasmic domains and talin form a platform for the recruitment of other focal-adhesion proteins. Integrin-linked kinase (ILK), and isoforms of particularly interesting Cys-His-rich protein (PINCH) and parvin form the IPP complex (Fig. 2) in the cytoplasm, and this complex is recruited to focal adhesions through interactions with other factors, such as paxillin. Other proteins such as vinculin and focal adhesion kinase (FAK) are recruited to the nascent focal complex in a sequential manner121. The maturation of focal adhesions involves clustering of active, ligand-bound integrins and the assembly of a multiprotein complex that is capable of linking integrins to the actin cytoskeleton and communicating with signalling pathways.
Among the integrins, 1 integrin contributes to a large number of integrin heterodimers whereas 3 integrin has an important adhesion role in platelets — an excellent system in which to study cell–ECM adhesion1. 1 and 3 integrins are widely expressed, and studies on the function of 1 and 3 integrins have provided many general insights into integrin-mediated adhesion. Deletion of the highly conserved 1 integrin gene in different organisms has been associated with defects in adhesion, proliferation, survival and polarity2, 3, 4, 5, 6. However, although these experiments showed that integrins are important for these processes, they provided few insights into how these processes are regulated. Deletions of certain focal-adhesion molecules in different organisms display strikingly similar phenotypes to the 1-integrin-null phenotype, which indicates that certain intracellular proteins might have key roles in the regulation of the function of 1 integrins.
Three proteins that have emerged from these studies as important regulators of integrin-mediated signalling are the integrin-linked kinase (ILK), and the adaptor proteins PINCH (particularly interesting Cys-His-rich protein) and parvin. These molecules form a heterotrimeric complex we refer to as the IPP complex, which is named after its components in order of their discovery. Recent reports have provided a wealth of data to expand the known functions of the IPP complex into almost every aspect of cell behaviour and fate. This review will provide an overview of our current knowledge regarding the function of the IPP complex. Interactions between the IPP components and numerous binding partners will be discussed to explain how the IPP complex functions both as an adaptor between integrins and the actin cytoskeleton, and as a hub that regulates several signalling pathways. Furthermore, this review will address the latest results in the ongoing controversy regarding the function of the putative kinase activity of ILK. We will conclude by describing the results of in vivo studies in model organisms, which provide insights into the role of the IPP-mediated integrin-signalling functions during development.
Identification, architecture and assembly of IPP
ILK was identified in 1996 in a yeast two-hybrid screen for proteins that could bind to the cytoplasmic tail of 1 integrin7. The protein that was cloned contained three domains (Fig. 2). The N-terminal domain contains three ankyrin repeats, which mediate protein–protein interactions, and a putative fourth ankyrin repeat that lacks conserved residues. The C terminus shares significant sequence homology to Ser/Thr protein kinases. A putative pleckstrin homology (PH) domain is situated between these two domains and partially overlaps them. Cell-culture experiments indicated that the physiological ligand of the ILK PH domain is phosphatidylinositol-3,4,5-trisphosphate (PtdIns(3,4,5)P3)8, 9. ILK is the central component of the IPP complex; it binds PINCH proteins through the N-terminal ankyrin-repeat domain and parvins through the kinase domain. It also links the complex to the cytoplasmic tails of 1 and 3 integrins7, 10, 11, 12, 13, 14, but it is not known whether it binds to other integrins. A single ILK isoform has been identified in all species discussed in the following sections.
Figure 2 | Anatomy of the IPP complex and its binding partners.
Integrin-linked kinase (ILK) consists of three domains, N-terminal ankyrin (ANK) repeats, a plekstrin homology (PH) domain and a C-terminal kinase domain. ANK1 binds to the LIM1 domain of particularly interesting Cys-His-rich protein (PINCH) isoforms as well as to the ILK-associated phosphatase (ILKAP). The PH domain probably binds to phosphatidylinositol-3,4,5-trisphosphate (PtdIns(3,4,5)P3). The kinase domain of ILK binds to parvins, paxillin, MIG2/kindlin-2, the cytoplasmic tails of integrins, the kinase substrate AKT/PKB (protein kinase B) and the kinase phosphatidylinositol-3-kinase-dependent kinase-1 (PDK1). PINCH isoforms, which contain five LIM domains, bind to receptor tyrosine kinases (RTKs) through the Src-homology-2 (SH2)–SH3 adaptor NCK2, thereby coupling growth-factor signalling to integrin signalling. PINCH1 binds to Ras suppressor-1 (RSU1) and thymosin-4 (T4) to influence Jun N-terminal kinase (JNK) signalling and cell migration/survival, respectively. - and -parvins can bind to F-actin directly, as well as indirectly through binding to paxillin or HIC5 (-parvin only) or -actinin (-parvin only). -Parvin also binds to the Ser/Thr kinase testicular protein kinase-1 (TESK1), whereas -parvin binds to the guanine nucleotide-exchange factor -PIX, which influences actin remodelling through the GTPases Rac and Cdc42. Interactions with integrins and the cytoskeleton also occur through a MIG2/kindlin-2–migfilin–filamin complex.
PINCH1 (also known as LIMS1) was originally identified in 1994 as a marker for senescent erythrocytes, and was shown to bind to ILK in 1999 (Refs 15,16). A second isoform, PINCH2 (also known as LIMS2), was predicted by sequence-database mining and was subsequently characterized17, 18. PINCH1 and PINCH2 are adaptor proteins that consist of five LIM domains and tandem nuclear localization signals17, 19, and both isoforms bind to ILK through the N-terminal LIM domain in a mutually exclusive manner10, 18. There is extensive overlap in the expression patterns of PINCH1 and PINCH2 in adult tissues, and both genes are expressed in smooth-muscle layers of the developing embryo19.
Parvins — a family of proteins that consists of actopaxin/CH-ILKBP/-parvin, affixin/-parvin and -parvin — bind to ILK through the second of two calponin homology (CH) domains13, 14, 20, 21. The interaction between -parvin and ILK is partially dependent on PtdIns(3,4,5)P3 (Ref. 22) and on phosphorylation of -parvin by CDC2 and mitogen-activated protein kinase (MAPK)23, 24, 25. It is unclear whether the other parvins are regulated in a similar manner, although -parvin can also be phosphorylated14 (see also Table 1). - and -parvin have overlapping expression patterns in various tissues, but the expression of -parvin is restricted to the haematopoietic system26. It is therefore possible that different IPP complexes can assemble within the same cell. This idea was further supported by a recent study showing that small inhibitory RNA (siRNA)-mediated knockdown of -parvin in HeLa cells stimulated Rac activity and lamellipodium formation. These findings indicate that -parvin can function as a negative regulator of Rac in cells that express both - and -parvin27. Overexpression of -parvin in HeLa cells promoted apoptosis, perhaps because of competition with -parvin for binding to ILK. Monitoring the expression levels of parvins in cells that are exposed to specific conditions, and immunoprecipitation of discrete IPP complexes will determine whether this data is biologically relevant, and will enable the study of the mechanisms that regulate the expression of the different parvin isoforms.
IPP-complex assembly and stability. The assembly of the IPP-complex precedes cell adhesion, which indicates that these complexes first form in the cytosol, independently of adhesion signals28. The stability of the individual IPP components is dependent on complex formation, because RNA interference (RNAi)-mediated depletion of one member of the complex results in degradation of the other components by a proteasome-mediated process29. This complicates the interpretation of results from genetic-deletion or RNAi-depletion studies because the defects that are associated with deletion of one component might be due to diminished levels of other components. Recent work is beginning to address the specific roles of IPP-complex members, taking into account their mutual stabilization and degradation30. Degradation of ILK or the PINCH proteins can be prevented by expression of the PINCH1 or PINCH2 LIM1 domain in Pinch1-deficient cells31 or expression of the ILK ankyrin repeats in Ilk-deficient cells, respectively (C. Grashoff and R.F., unpublished data), but only a complex that consists of full-length proteins efficiently assembles into focal adhesions in vertebrate cell culture20, 28, 32. Although the full-length IPP complex is required for focal-adhesion assembly in vertebrates, other factors are also required to assemble the IPP complex into focal adhesions, including the adaptor molecule paxillin, and possibly MIG2/kindlin-2 (Refs 28,33). Both of these proteins directly bind to the IPP complex through the kinase domain of ILK33, 34. Many more proteins might also be involved at this stage of IPP-complex assembly and targeting. Interestingly, in Drosophila melanogaster embryos, the PINCH homologue, steamer duck (STCK), is not required to localize ILK to sites of cell adhesion35.
IPP-binding partners
There are many molecules that have been shown to interact with the IPP complex36. The function of the IPP complex as a signalling platform is achieved through its direct interaction with factors that function as upstream regulators of many different signalling pathways (Fig. 3). This section will summarize the known binding partners of the IPP complex and will provide examples where signalling specificity may be achieved through differential binding of molecules to PINCH and parvin isoforms.
Figure 3 | Signalling through the IPP complex.
The integrin-linked kinase (ILK), particularly interesting Cys-His-rich protein (PINCH), parvin (IPP) complex has been implicated in the control of signalling pathways through both phosphorylation of downstream targets (most notably AKT/protein kinase B (PKB) and glycogen-synthase kinase-3 (GSK3) and binding to upstream effectors of the Jun N-terminal kinase (JNK) signalling pathway and regulators of small-molecular-weight GTPases. The activity of the complex is upregulated by phosphatidylinositol 3-kinase (PI3K) and downregulated by the phosphatases ILK-associated protein (ILKAP) and phosphatase and tensin homologue deleted in chromosome 10 (PTEN). Growth-factor-mediated signalling through receptor tyrosine kinases (RTKs) might be transduced to the IPP complex through the receptor-tyrosine-kinase adaptor protein NCK2. The signalling pathways that are shown are limited to those that have been experimentally described to be influenced by the IPP complex. AP-1, activator protein-1; BAD, BCL2-antagonist of cell death; COX2, cyclooxygenase-2; CREB, cAMP-response-element-binding protein; ECM, extracellular matrix; HIF1, hypoxia-inducible factor-1; iNOS, inducible nitric-oxide synthase; LEF/TCF, lymphoid enhancer factor/T-cell factor; MAP4K, mitogen-activated-protein-kinase-kinase-kinase kinase; MLC, myosin light chain; MMP9, matrix metalloprotease-9; mTOR, mammalian target of rapamycin; -NAC, nascent polypeptide-associated complex and co-activator-; NF-kB, nuclear factor-kB; p70S6K, p70 ribosomal S6 kinase; -PIX, activating PAK-interactive exchange factor-; PtdIns(3,4,5)P3, phosphatidylinositol-3,4,5-trisphosphate; RSU1, Ras suppressor-1; VEGF, vascular endothelial growth factor.
ILK-interacting partners. As mentioned above, vertebrate ILK can bind directly to the cytoplasmic tails of 1 and 3 integrins7, 11, 12, and it is indirectly connected to the actin cytoskeleton through its interaction with parvins (see below). Interactions with the cytoskeleton can also occur through the LD motif and LIM-domain adaptor protein paxillin, which binds to F-actin through interactions with -parvin and the actin-binding adaptor molecule vinculin20, 37, 38. Paxillin binds to ILK through a paxillin-binding site (PBS) within the kinase domain of ILK33. Furthermore, Caenorhabditis elegans ILK binds to UNC-112, the orthologue of vertebrate MIG2/kindlin-2 (Ref. 34). In vertebrates, kindlin-1 binds to the cytoplasmic domains of 1 and 3 integrins39, and its loss causes Kindler syndrome40. MIG2/kindlin-2 binds to migfilin, which in turn binds to filamin — an adaptor protein that interacts with several molecules, including filamentous (F)-actin and integrins41, 42 — and provides another connection between ILK and the actin cytoskeleton.
Binding partners of the PINCH isoforms. The signalling specificity of the IPP complexes depends on the presence of the different PINCH or parvin isoforms. When PINCH2 is overexpressed in a basal PINCH1-expressing background, it competes with PINCH1 for binding to ILK but cannot transduce integrin-mediated signals that control cell spreading and migration18. Also, although the expression of a chimeric PINCH that consists of the PINCH1 LIM domains and the PINCH2 C-terminal tail cannot restore spreading in PINCH1-knockdown HeLa cells30, expression of full-length PINCH2 in a Pinch1-null background completely restores the adhesion and spreading defects of Pinch1-null fibroblasts31. Therefore, accessory proteins that differentially bind to PINCH isoforms are probably responsible for transducing the signals that control cell spreading.
Functional differences between PINCH1 and PINCH2 might arise from differential binding of the Ras-suppressor protein RSU1. RSU1 has been shown to function as a negative regulator of growth-factor-induced Jun N-terminal kinase (JNK) activation43, 44, 45, 46 (Fig. 3). RSU1 is also known to interact with the LIM5 domain of PINCH1 in D. melanogaster45 and vertebrates46, but this interaction is specific for PINCH1. PINCH2 does not bind RSU1, because the sequence of the LIM5 domain is different19.
Thymosin-4 binds to LIM domains -4 and -5 of PINCH1, upregulates ILK activity and positively influences migration and survival of cardiac cells47. Whether PINCH2 and thymosin-4 interact is not known, but a conditional knockout of Pinch1 in murine ventricular cardiomyocytes shows no discernable phenotype48, indicating that PINCH2, which is also expressed in the heart19, might compensate for the loss of PINCH1.
PINCH1 also binds to the receptor-tyrosine-kinase adaptor protein NCK2 in vitro through a LIM4–SH3-domain (Src-homology-3 domain) interaction49, 50. Mutations in vertebrate PINCH1 that disrupt PINCH1–NCK2 binding reduce the amount of PINCH1 that is found in focal adhesions49, but the relevance of a PINCH1–NCK2 interaction is not clear. There is no evidence for such an interaction in C. elegans or D. melanogaster. Mice that carry a Nck1 or Nck2 genetic deletion, are phenotypically normal, whereas mice that lack both proteins die during embryogenesis. The migration and cytoskeletal defects of Nck1-/- Nck2-/- fibroblasts51 are rescued when NCK1 is reintroduced, although the PINCH1–NCK interaction is specific for NCK2. Furthermore, NCK1 has been demonstrated not to bind to PINCH1 (Ref. 52). These results indicate that the interaction between PINCH1 and NCK2 might not be essential in vivo.
Parvins have overlapping binding partners. -Parvin binds to F-actin directly, as well as indirectly through an interaction with paxillin20. It is not yet known whether ILK and a parvin isoform bind to the same molecule of paxillin, or if two paxillin molecules bind to a single IPP complex. HIC5, a paxillin-related protein, also binds to -parvin20. HIC5 binds to many of the proteins that paxillin binds to, but it also shuttles to the nucleus, where it modulates the expression of several genes53, 54. In addition, -parvin specifically binds to TESK1, a Ser/Thr kinase that phosphorylates the actin-regulating protein cofilin55. -Parvin binds to F-actin, but can also bind to the actin-crosslinking protein -actinin56 and the guanine nucleotide-exchange factor -PIX57. It therefore provides a connection between the IPP complex and the actin-regulating GTPases Rac1 and Cdc42 (Fig. 3). -PIX, in turn, binds to PAK1 (Ref. 58), a Rac1/Cdc42 effector that regulates cytoskeletal dynamics through the LIM-kinase–ADF–cofilin pathway (where ADF stands for actin-depolymerizing factor)59. Furthermore, -PIX binds to the protease subunit calpain-4 (Ref. 60). Calpain-4 has been shown to cleave talin, and this cleavage is the rate-limiting step in the disassembly of focal adhesions61. The functions of parvin-binding partners indicate that the parvins have a role in the regulation of actin dynamics and focal-adhesion turnover. -Parvin also binds to the membrane-repair protein dysferlin at the sarcolemma of skeletal muscle; an observation that reveals a potential role for -parvin in membrane repair62. Binding partners for the less-well-characterized -parvin have yet to be identified.
Other roles of the IPP-complex proteins? Deletion of one component of the IPP complex results in degradation of the other components, but this degradation is not complete29, 63, which indicates that the individual components might have extra roles outside the complex. PINCH1 shuttles between the nucleus and the cytoplasm in Schwann cells, and it has been found in the nucleus of the mouse primitive endoderm, and muscle cells of C. elegans. These observations indicate that PINCH1 might have a role in gene regulation or signalling between the cytoplasm and the nucleus17, 63, 64. -Parvin localizes to the sarcolemma in skeletal muscle, where it might have a role in Ca2+-dependent membrane repair, together with dysferlin62. ILK redistributes to cell–cell contacts in differentiating keratinocytes, where it participates in the early steps of adherens-junction formation65, 66.
ILK catalytic activity
ILK has been shown to function as an adaptor protein, but recent reports have indicated that ILK also has catalytic activity. The kinase domain of ILK shows significant homology to Ser/Thr protein kinases, with the exception of sequences within the catalytic loop and the conserved DXG motif67, 68. These differences from the canonical kinase sequence are difficult to reconcile with the observed kinase activity, because ILK lacks an obvious catalytic base and Mg2+-chelating residues. Furthermore, the sequence of the ILK catalytic domain is divergent across different species69 (Fig. 4). Such a tolerance for mutation indicates that if ILK possesses kinase activity, it is probably unnecessary for its function. Nevertheless, recombinant, purified ILK has been shown to phosphorylate several substrates in vitro7, 70, 71 (Table 1), and it is possible that ILK retains residual kinase activity that is readily detected in vitro. Because kinetic rates of the kinase reaction have not been reported and the evidence for in vivo kinase activity is weak, it is not known whether ILK possesses sufficient activity to function as a physiologically relevant kinase in vivo.
Figure 4 | Divergence of the kinase domain of ILK.
Alignment of integrin-linked kinases (ILKs) from several species as well as the related kinase transforming-growth-factor--activated kinase-1 (TAK1) with conserved subdomains of protein kinases, numbered according to Hanks et al.67, 68. The subdomain X is poorly conserved and contains no consensus amino acids, so it was not included. Conserved domains that are significantly altered in ILK are shown in red. One of three Gly residues in subdomain I is conserved in all ILK proteins. A Lys in subdomain II that is required for the phosphotransfer reaction82 is also conserved as a basic residue. However, the catalytic Asp residue in subdomain VI and the DFG sequence in subdomain VII that is required to align the -phosphate of ATP, are both missing. The conserved Lys, which neutralizes the charge on the -phosphate of ATP, and the conserved Asn, which chelates the secondary Mg2+ are also missing. All active protein kinases contain a conserved Asp in regions VI and VII except for the haspins — a unique family of histone kinases that have a role in mitosis122, and lack the conserved Asp in subdomain VII but contain the catalytic Asp in subdomain VI123. ILK is the only protein kinase that is known to be missing both residues. C. elegans, Caenorhabditis elegans; D. melanogaster, Drosophila melanogaster; M. musculus, Mus musculus; X. laevis, Xenopus laevis.
ILK model substrates. The most widely used readouts for ILK activity are the phosphorylation of GSK3 and AKT/PKB (protein kinase B), which regulate many different signalling pathways (Fig. 3). AKT/PKB activation requires phosphorylation of Thr at position 308 by phosphatidylinositol 3-kinase (PI3K)-dependent kinase-1 (PDK1) and Ser at position 473 by PDK2 (which is more appropriately known as hydrophobic-motif kinase (HMK)). The identification of PDK1 is unambiguous, but the identity of PDK2/HMK continues to be debated. ILK, among other candidates (Box 1), has been proposed to function as a HMK and this idea is further supported by immunoprecipitation assays showing that ILK directly binds to AKT/PKB71.
Regulation of ILK activity. ILK-dependent phosphorylation is regulated in a PI3K-dependent manner. Inhibitors of PI3K activity reduce ILK activity in cell-lysate immunoprecipitates and impair the phosphorylation of putative ILK substrates in cell culture, whereas overexpression of the PI3K catalytic subunit or the addition of PtdIns(3,4,5)P3 increase ILK-dependent kinase activity in cell culture and in vitro, respectively8. The expression of thymosin-4 in cardiomyocytes also increases the activity of ILK, as measured by phosphorylation of AKT/PKB on Ser47347. Conversely, the catalytic activity of ILK is negatively regulated by the phosphatase ILK-associated protein (ILKAP). ILKAP reduces the kinase activity of ILK in vitro and the phosphorylation of GSK3 in vivo, but the phosphorylation of AKT/PKB remains unaffected72, 73. This demonstrates that the mechanisms of GSK3 and AKT/PKB activation are different, despite the fact that both proteins have been described as ILK substrates. The mechanism by which ILKAP reduces ILK activity towards certain substrates is currently unknown. It has also been reported that ILK can autophosphorylate in vitro2, but the function of autophosphorylation of ILK in vivo, if present at all, remains to be investigated.
Perturbation of ILK function. Despite evidence in favour of ILK having a kinase function, genetic experiments have failed to demonstrate a crucial role for the kinase activity of ILK in several cell types. Alterations in the phosphorylation status of AKT/PKB or GSK3 have not been observed in Ilk-/- fibroblasts74. Furthermore, the phosphorylation status of putative ILK substrates is unchanged in chondrocytes or keratinocytes that do not express ILK (K. Lorenz and R.F., unpublished data, and Ref. 75). By contrast, deletion of Ilk in immortalized macrophages results in diminished phosphorylation of both AKT/PKB and GSK3, which is reversed by transfection of wild-type ILK, but not a kinase-deficient ILK mutant76. This discrepancy might reflect cell-type-specific differences, or it might reveal an increased dependence on the kinase activity of ILK as a result of immortalization.
Small-molecule inhibitors that were designed to specifically inhibit the kinase activity of ILK, prevent phosphorylation of AKT/PKB and GSK3 when introduced in ILK-overexpressing cell lines9, 71, 77. Further characterization of one of these inhibitors, KP-392, revealed that it disrupts the interaction between ILK and both -parvin and paxillin, and blocks the accumulation of -parvin and paxillin into focal adhesions22.
Similarly, several different point mutations in ILK, which were designed to abolish catalytic activity, disrupt protein–protein interactions and might function by preventing the assembly of functional IPP complexes. Overexpression of wild-type ILK in cell-culture experiments results in increased phosphorylation of AKT/PKB and GSK3 in many cell types. Conversely, overexpression of a dominant-negative mutant of ILK (Glu359Lys) reduces phosphorylation of AKT/PKB and GSK3. Although the Glu359Lys mutation was reported to impair the kinase activity of ILK8, 78, other studies conclude that this mutation has no effect on the kinase activity of ILK in vitro14, 79. Instead, the Glu359Lys mutation disrupts the interaction between ILK and both -parvin and paxillin, which results in a failure to assemble ILK into focal adhesions14, 79. Several inactivating point mutations in the kinase domain of ILK have been used to dissect the kinase function, but these mutations (Arg211Ala, Ser343Ala, and Lys220Ala) also disrupt ILK–protein interactions, which makes it impossible to derive interpretations that are based solely on the lack of kinase activity. The Arg211Ala mutation results in deficient binding to -parvin22, and this interaction is required to target ILK to focal adhesions28. The Ser343Ala mutation abolishes binding to AKT/PKB, but this mutant protein localizes to focal adhesions71, 80. The Lys220Ala mutant fails to bind to -parvin81. Another mutation at this site, Lys220Met binds to -parvin but shows no catalytic activity14, but a second activating mutation, Ser343Asp, reverses this defect69 despite a requirement of Lys220 for protein-kinase activity68, 82. The Phe438Ala mutant, which does not assemble into focal adhesions, has been proposed not to bind to MIG2/kindlin-2 (Ref. 28). Furthermore, mutations in ILK that show dominant-negative effects in cell culture (Glu359Lys or Lys220Met), or abolish the activity of the protein kinase Raf at non-permissive temperatures (Pro358Ser), completely rescue the phenotype of the ILK loss-of-function mutations in C. elegans (only Glu359Lys was tested) and D. melanogaster (all three ILK mutants). This indicates that kinase activity is dispensable for ILK function in these organisms34, 83. However, residual kinase activity might be sufficient to completely restore a wild-type phenotype.
RNAi studies in transformed cells demonstrate that knockdown of ILK, PINCH1 or -parvin protein expression correlates with reduced phosphorylation of AKT/PKB Ser473 (Refs 29,76,84–86). Furthermore, AKT/PKB Thr308 phosphorylation is also reduced in the absence of PINCH1 (Ref.29). AKT/PKB does not localize to the plasma membrane in -parvin-depleted HeLa cells, despite the presence of ILK in focal adhesions, which demonstrates that -parvin is required for the correct targeting of AKT/PKB before activation84. However, if AKT/PKB is constitutively targeted to the membrane, phosphorylation of Ser473 does not depend on ILK29, 84. It should be noted that the increased apoptosis that results from knockdown of ILK or PINCH1 is not reversed on restoration of AKT/PKB phosphorylation, which indicates that these molecules regulate numerous survival pathways.
Taken together, the data from mutagenesis, small-molecule inhibition and RNAi all indicate a model whereby ILK activates AKT/PKB indirectly by facilitating the translocation of AKT/PKB to the plasma membrane in an -parvin-dependent manner. Once AKT/PKB is located at the plasma membrane, it can be phosphorylated by other kinases.
The role of the IPP complex in invertebrates
To study the contribution of the IPP complex to development, genetic deletion of most components has been achieved in C. elegans, D. melanogaster and mice (see Supplementary information S1 (table)). The phenotypes caused by these deletions confirm a role for the IPP complex as an adaptor complex between the ECM and the actin cytoskeleton, but insights into the signalling function are not as forthcoming, and a role for the ILK catalytic activity has not been established. Differences in phenotypes among IPP components, particularly in mice, reveal that, in addition to a common function as part of the IPP complex, each component has discrete functions.
Invertebrates represent relatively simple systems to study integrin-mediated functions. C. elegans has one integrin (PAT-3) and two integrin (PAT-2 and INA-1) subunits; D. melanogaster has two integrin (PS and v) and five integrin (PS1–5) subunits. Both organisms have a single orthologue of ILK, PINCH and parvin, which allows straightforward analysis of deletion phenotypes. C. elegans has a second PINCH1-related gene but this shows only 34% identity with PINCH1 and contains a putative endoplasmic reticulum (ER) targeting sequence.
Studies in C. elegans. In C. elegans, genetic deletion of pat-3 (Ref. 3), pat-4 (which encodes the ILK orthologue)34, unc-97 (which encodes the PINCH orthologue)17 or pat-6 (which encodes the parvin orthologue)87 all produce mutants with a similar PAT phenotype. Although a direct physical interaction between PAT-3 and PAT-4 has not been reported, deletion of pat-3 or the ECM component unc-52 (which encodes the orthologue of perlecan) results in mislocalization of PAT-4, which indicates that PAT-4 requires integrins and ECM to localize to dense bodies34. Recruitment of PAT-4 to dense bodies, and the correct organization of PAT-3 at the cell membrane also depend on UNC-112 (the orthologue of MIG2/kindlin-2)34, 88. However, there is no evidence that PAT-4 regulates the GSK3 signalling pathway in this organism, as the defects that arise from deletion of pat-4 are distinct from the phenotype of the GSK-3 mutant in C. elegans34. A novel Zn2+-finger protein that is unique to C. elegans, UNC-98, binds to UNC-97 at M-lines and dense bodies89. Both proteins are also detected in the nucleus, which indicates that they might be involved in gene regulation17, 89.
Studies in D. melanogaster. In D. melanogaster, deletion of PS, the orthologue of the vertebrate 1 integrin subunit2, or loss of function of ILK83 or the PINCH orthologue35, result in striking embryonic muscle-attachment defects. PS and PINCH-deficient flies also show dorsal-closure defects, which indicates that cell migration is impaired2, 45. Adult chimeric flies display a blister phenotype in wing regions that lack these proteins. These phenotypes are consistent with the loss of cell adhesion to the ECM. However, PS mutants show adhesion defects that result from the detachment of the cell membrane from the ECM, whereas the ILK and PINCH mutants are characterized by the detachment of the actin cytoskeleton from the cell membrane. Interestingly, ILK localizes normally to muscle-attachment sites in PINCH-mutant embryos, which indicates that the PINCH loss-of-function phenotype is not caused by the mislocalization of ILK35. The loss-of-function mutations in PINCH affect the LIM domains 3–5, therefore it is possible that the LIM domains 1 and 2 are still expressed, and are sufficient to retain correct ILK localization in this organism.
Cytoskeletal reorganization and changes in cellular shape that are required during dorsal closure are coordinated by the JNK signalling pathway (for a review see Ref. 90). Therefore, the dorsal-closure defect in PINCH-mutant flies might be explained, in part, by disrupted JNK signalling. PINCH might regulate JNK signalling through two distinct pathways. The first involves the interaction between PINCH and RSU1 (Fig. 3). Both PINCH and RSU1 genetically interact with Misshapen, a MAPK-kinase-kinase kinase (MAP4K) in the JNK signalling pathway, which demonstrates that both proteins can negatively regulate this pathway45. The expression and/or stability of PINCH and RSU1 are mutually dependent on one another, so the relative contributions of PINCH and RSU1 in modulating JNK signalling have not been determined. The second and more-speculative mechanism involves a PINCH–Dreadlocks–Misshapen pathway. Dreadlocks, the D. melanogaster orthologue of NCK2, interacts with Misshapen91. Although mammalian PINCH1 binds to NCK2 in vitro52, an interaction between Dreadlocks and PINCH has not been reported. However, the affinity between mammalian NCK2 and PINCH1 is weak in vitro, and it is possible that it could not be detected.
The two main pathways that regulate survival and proliferation involve the phosphorylation of AKT/PKB and the stabilization and nuclear translocation of -catenin as a consequence of GSK3 phosphorylation (Fig. 3). However, there is no evidence that ILK is involved in these pathways in D. melanogaster as ILK loss-of-function produces a different phenotype to loss of -catenin or AKT/PKB34, 83, 92, 93. Furthermore, overexpression of ILK does not affect signalling through -catenin83.
Functions of IPP in mammalian systems
The biological functions of the members of the IPP complex have been examined in several cell types. Many IPP functions can be reconciled with a role for the IPP complex at focal adhesions. For example, the regulation of podocyte adhesion and spreading23, endothelial cell and cardiomyocyte migration47, 85, platelet aggregation11, 12, neuronal spreading and outgrowth77, 94, and leukocyte recruitment95 reveal a role for the IPP complex in actin-cytoskeleton dynamics and integrin activation.
Studies in mice. Loss of expression of 1 integrin4, 5, ILK74, PINCH1 (Refs 48,63) or -parvin (H. Chu and R.F., unpublished data) all result in embryonic lethality. However, these mutants show subtle differences in their phenotypes, which indicates that distinct defects might underlie these phenotypes. 1 Integrin-/-, Ilk-/- and Pinch1-/- embryos arrest at the peri-implantation stage of embryonic development, but the Pinch1-/- embryos die at the embryonic day (E)6.5–E7.5 whereas the 1 integrin-/- and Ilk-/- embryos die at E5.5–E6.5. -Parvin-/- embryos develop the furthest, but die following implantation. The temporal differences in embryonic lethality between 1 integrin-/- or Ilk-/- embryos and Pinch1-/- embryos are unlikely to be due to PINCH2 compensation, because expression of PINCH2 is not detected until later stages of development19. On the other hand, Pinch2-/- mice are viable, probably as a result of PINCH1 compensation31. It will be important to determine whether the relatively long survival of -parvin-/- embryos is the result of compensation by other members of the parvin family and to investigate whether parvins might have a crucial function either before, or during, peri-implantation development.
Embryonic lethality in Pinch1-null mice is associated with an increase in endodermal-cell apoptosis, which demonstrates that PINCH1 regulates cell survival63. This observation is further supported by PINCH1-knockdown data in HeLa cells29. However, it remains to be determined whether AKT/PKB is involved in the PINCH1-mediated apoptotic pathway in embryos or if, as in HeLa cells, an AKT/PKB-independent pathway exists.
Conditional knockouts of IPP members have been created to overcome the difficulties associated with early embryonic lethality. Mice that carry conditional deletions of 1 integrin or Ilk in chondrocytes show skeletal defects75, 96, 97 and develop chondrodysplasia. In culture, 1 integrin- and Ilk-mutant chondrocytes had defective spreading, abnormal F-actin distribution, and impaired adhesion, although the Ilk-mutant chondrocytes had a less-severe adhesion defect. Therefore, deletion of either protein impairs the connection between the ECM and the actin cytoskeleton that is required for cell spreading. In addition, reduced proliferation of 1 integrin- or Ilk-deficient chondrocytes is associated with a defect in the G1–S transition, but impaired cytokinesis in 1 integrin-/- chondrocytes reveals a defect in the G2–M transition.
Conditional deletion of Ilk in endothelial cells results in impaired vascular development and embryonic lethality98. Integrin activation is reduced in the absence of ILK, which indicates a migration defect, and there is an increase in apoptosis concomitant with decreased AKT/PKB phosphorylation on Ser473. However, apoptosis is reduced when ILK is reintroduced, but not when constitutively active AKT/PKB is introduced. This indicates that, as in HeLa cells, an ILK-dependent, AKT/PKB-independent pathway might operate in endothelial cells to regulate apoptosis.
Immortalized Ilk-/- macrophages also show decreased AKT/PKB phosphorylation on Ser47376. In addition, inhibition of ILK or PI3K activity, expression of dominant-negative ILK and knockdown of ILK protein levels revealed a role for the IPP complex in AKT/PKB- and GSK3-mediated signalling in neurons and leukocytes77, 94, 95, 99, 100. However, fibroblasts and chondrocytes that are derived from Ilk-/- mice do not show altered phosphorylation of AKT/PKB74, 75, 97. These discrepancies might reflect cell-type-specific differences in the requirement for ILK involvement upstream of AKT/PKB and GSK3.
Insights from studies using embryonic bodies. Because mice that lack 1 integrin, ILK or PINCH1 die early in development, embryoid bodies were used to identify the specific defects that cause peri-implantation lethality (Box 2). 1 Integrin-/- embryoid bodies fail to deposit a basement membrane due to a defect in the synthesis of laminin101, 102. By contrast, Ilk-/- (Box 2, figure, part c) and Pinch1-/- (Box 2, figure, part d) embryoid bodies produce a basement membrane but have defects in epiblast polarization and cavity formation; these defects are less severe in Pinch1-/- embryoid bodies, consistent with the longer survival time in vivo63, 74. The defects in epiblast polarization are accompanied by abnormal localization of F-actin. These studies provide evidence that 1 integrins have separate functions from ILK and PINCH1 at the peri-implantation stage and highlight an important role for the IPP complex in organizing the actin cytoskeleton. Evidence also indicates that the IPP complex could regulate the actin cytoskeleton indirectly through NCK2, which binds to the actin modulators WASP (Wiskott–Aldrich syndrome protein), PAK (p21-activated kinase) or DOCK180 (180-kDa protein downstream of CRK)103. In addition, Pinch1-/- embryoid bodies show abnormal cell–cell adhesion and impaired endoderm survival63.
A reduction of ILK protein levels in Pinch1-/- embryoid bodies, similar to that observed in Pinch1-siRNA-treated HeLa cells29, provides a possible explanation for the phenotypic similarities between Ilk-/- and Pinch1-/- embryoid bodies63. However, the defects that are observed in Pinch1-/- embryoid bodies represent ILK-independent functions and indicate that the functional interdependence between ILK and PINCH1 is not complete. This is supported by the observation that Ilk-/- cells still express low levels of PINCH1, and vice versa31, revealing the existence of different functional pools of these proteins: a main pool in which ILK and PINCH1 are mutually dependent, and a second, smaller pool in which these proteins function independently.
Recent data have also shown a role for IPP members in the regulation of cell polarity and cell–cell contacts63, 74, 104. Cell–cell adhesion defects in Pinch1-/- embryoid bodies are associated with a diffuse distribution of E-cadherin and the absence of adherens junctions63. However, how PINCH1 regulates cell–cell adhesion is poorly understood. Localization of PINCH1 to cell–cell contacts has not been reported so far, but it is possible that PINCH1 regulates cell–cell adhesion in an indirect manner. Although a role for ILK in cell–cell contacts has been suggested based on studies in keratinocytes65, 66, Ilk-/- embryoid bodies do not show cell–cell adhesion defects. Therefore, PINCH1-mediated regulation of cell–cell adhesion must be ILK-independent. Unlike keratinocytes66 and epiblast cells, trophoectodermal cells or primitive endoderm cells of embryoid bodies can polarize in the absence of ILK74, which indicates that different epithelial cells might have different requirements for ILK.
Conclusions
The IPP complex serves as an important transducer of extracellular signals to control many aspects of cell morphology and cell behaviour. Although there are many functions that are likely to be common among all cell types, certain functions seem to be cell-type specific. How does the IPP complex achieve specificity? The answer is not clear, but differential binding of ILK to PINCH and parvin isoforms is likely to be involved. Titration of binding sites by one isoform might lead to trans-dominant effects on other isoforms. Therefore, understanding how cells that express numerous isoforms of PINCH and parvin control the composition of the IPP complex to achieve spatial and temporal control over IPP function will be an important direction for future investigations.
In addition, the role of ILK as a kinase that is involved in the phosphorylation of AKT/PKB and GSK3 is still unclear. Immortalized cell lines or overexpressed ILK have been used in most of the studies, and few reports have examined the role of ILK in primary cells. Many of the effects of ILK-specific inhibitors or PI3K inhibitors that are observed in ILK-overexpressing cell lines do not manifest in cells that express basal levels of ILK. For example, cyclin-D1-promoter activity in epithelial cells is reduced in response to PI3K inhibitors only when ILK is overexpressed, which indicates that PI3K-mediated pathways are not involved in the basal control of cyclin-D1 expression105. On the other hand, cyclin-D1 expression in endothelial cells is upregulated in the presence of vascular endothelial growth factor (VEGF) in an ILK-dependent manner85, and peroxisome-proliferator-activated receptor- (PPAR)-dependent upregulation of ILK expression in keratinocytes correlates with increased phosphorylation of AKT/PKB on Ser473 (Refs 106,107). Interestingly, ILK expression is upregulated in many human tumours108, 109, 110, and in vivo models have revealed that ILK overexpression leads to the development of mammary tumours in mice111. Differentiating the functions of ILK under normal and pathological conditions will provide insights into how ILK contributes to disease, and under which conditions ILK can be exploited therapeutically.
More fundamentally, careful enzymatic analysis of the kinase activity of ILK and determining the substrate specificity will address whether ILK is a biologically relevant AKT/PKB HMK. Furthermore, the elucidation of the three-dimensional structure of ILK is critical to adequately address the functions of ILK as a kinase and as an adaptor molecule. A structure will provide insights into how a kinase activity can be achieved in the absence of conserved subdomains, and will identify potential mechanisms whereby ILK activity is regulated by PtdIns(3,4,5)P3 and ILKAP. Extending a structural analysis to PINCH and parvin isoforms will allow us to identify the differences in their structures that lead to their distinct biological functions.
1 integrins.
