Defective granulation tissue formation in mice with specific ablation of integrin-linked kinase in fibroblasts – role of TGFβ1 levels and RhoA activity

The interaction of cells with the surrounding extracellular matrix (ECM) is a key regulatory mechanism controlling essential cellular functions and differentiation (Cukierman et al., 2002; Humphries et al., 2004; Larsen et al., 2006). Fibroblasts are resident cells in connective tissues of various organs and are among the cells most intimately surrounded and controlled by ECM (Grinnell, 2003). Their phenotype is regulated by the biochemical composition of the matrix environment through signals transduced by specific integrin receptors, which are activated in response to matrix recognition (Hynes, 2002). Fibroblast–ECM crosstalk is mediated predominantly by the β1 subfamily of integrin receptors. These receptors are non-covalently associated transmembrane heterodimers that share the β1 subunit, which associates with any of 12 different α subunits. In skin, collagens represent the most abundant group of ECM macromolecules and, for their recognition, fibroblasts are equipped with three different β1 integrins (α1β1, α2β1 and α11β1) (Popova et al., 2007; Zhang et al., 2006a), endowing cells with a set of tools for finely tuned cellular responses and adaptation to challenges from outside.

Fibroblasts orchestrate the reconstitution of connective tissues following injury by inducing ECM synthesis, deposition and subsequent contraction of granulation tissue. To fulfill these tasks, fibroblasts can differentiate to myofibroblasts, which represent a specialized phenotype that is characterized by the expression and assembly of α-smooth muscle actin (αSMA) into thick stress fiber bundles (Hinz, 2007). Myofibroblasts occur transiently in healing wounds and pathologically in fibrotic conditions, in which they contribute to impaired organ function due to tissue contractures (Desmouliere et al., 2005; Gabbiani, 2003). Fibroblast to myofibroblast conversion is a stepwise process that is strictly dependent upon the contact of fibroblasts with the surrounding ECM, the generation of mechanical forces across focal adhesion structures and the presence of transforming growth factor beta-1 (TGFβ1) (Hinz, 2007; Kessler et al., 2001; Tomasek et al., 2002). When repopulating damaged connective tissue, fibroblasts first acquire an activated, migratory phenotype with de novo formation of stress fibers composed of cytoplasmic actin. These cells produce and remodel collagen, thereby further increasing mechanical tension across the cell membrane. In response to high local levels of TGFβ1 and to increasing mechanical tension in the wound ECM, activated fibroblasts express αSMA, which is assembled into stress fibers (Hinz, 2006; Shephard et al., 2004). This leads to substantially augmented contractile activity of fibroblasts, which are then termed myofibroblasts, and to tissue contraction (Hinz et al., 2001).

In mammals, three TGFβ isoforms exist, with TGFβ1 being a potent positive regulator of myofibroblast conversion. They are synthesized as large precursor polypeptides with the latency-associated peptide (LAP) at the N terminus and the TGFβ peptide at the C terminus. In the endoplasmic reticulum, furin cleaves pro-TGFβ and induces formation of the small latent TGFβ complex, in which TGFβ remains non-covalently associated with its propeptide, LAP, and unable to bind to its receptors. The small latent complex might be further linked via LAP to latent TGFβ-binding proteins (LTBPs), giving rise to the large latent TGFβ complex. After secretion, the LTBPs target the large latent TGFβ complex to binding sites in the extracellular matrix, such as fibronectin or fibrillin-1. Thus, large amounts of TGFβ are stored in latent form in tissues (reviewed by Sheppard, 2006). Activation and release of bioactive TGFβ must be under tight control in view of its potent and pleiotropic effects. Different mechanisms have been described, including in vitro activation by extremes of temperature or pH. In vivo, thrombospondin-1 contributes to activation of TGFβ1 by binding to LAP and inducing a conformational rearrangement such that LAP cannot confer latency on the mature domain of TGFβ1 (Crawford et al., 1998; Murphy-Ullrich and Poczatek, 2000). Interestingly, integrins have turned out to be important players in latent TGFβ activation (Annes et al., 2004; Munger et al., 1999; Aluwihare et al., 2009; Mu et al., 2002; Sheppard, 2005).

In addition to the presence of TGFβ1, myofibroblast formation requires the transmission of mechanical stress to the fibroblast cytoskeleton across properly structured force-bearing matrix adhesions (Hinz, 2006). These consist of ECM ligands bound to integrins, which are organized into clusters, called focal adhesions (FAs), with the cytoplasmic tails of β1 subunits binding to a large submembranous complex of proteins containing talin, vinculin, paxillin, tensin, focal adhesion kinase, integrin-linked kinase (ILK) and many others (Cukierman et al., 2002; Hinz, 2006; Sawada and Sheetz, 2002; Thannickal et al., 2003). ILK is a scaffolding molecule that connects the intracellular domains of β1 and β3 integrin subunits to further bridging molecules (Hannigan et al., 1996; Nikolopoulos and Turner, 2001; Tu et al., 2001; Yamaji et al., 2001; Rosenberger et al., 2003). It supplies a structural platform linking ECM-bound integrins to actin cables and integrates several signaling pathways that regulate cell behavior (reviewed by Legate et al., 2006).

Actin organization and cell contractility are linked to transforming protein RhoA (Nobes and Hall, 1999; Chrzanowska-Wodnicka and Burridge, 1996). In muscle and non-muscle cells such as fibroblasts, RhoA signals via Rho-associated kinase (ROCK) to phosphorylate myosin light chain (MLC) and to inhibit MLC phosphatase. This is associated with increased cell contractility. Early phases of cell spreading are characterized by high activity of the RhoA-related GTPase Rac1, which facilitates spreading and lamellipodium formation, and concomitant suppression of RhoA activity by Rac1 (Sander et al., 1999).

Constitutive ablation of ILK function in mice results in peri-implantation lethality as a result of adhesion and polarization defects of the epiblast (Sakai et al., 2003). More recently, ILK inactivation has been targeted conditionally to selected cell types (Lorenz et al., 2007; Nakrieko et al., 2008; Gheyara et al., 2007; White et al., 2006; Grashoff et al., 2003). Common to all mutants, including the constitutively ablated early embryo, is the formation of sparse, short or condensed actin filaments, abnormal cell shape and dysregulated cell proliferation, which is either augmented or repressed, depending on the affected cell type.

We targeted ILK inactivation to fibroblasts in the skin of young mice to test the hypothesis that ILK maintains a crucial function as a docking structure in FAs connecting the ECM with the actin cytoskeleton. We report that myofibroblast formation is strongly impaired, leading to diminished dermal tissue repair in vivo. Primary ILK-deficient skin fibroblasts reveal an abnormal shape, enhanced RhoA activity, reduced TGFβ1 and αSMA production, and delayed directed migration. These defects were rescued by interfering with RhoA–ROCK signaling, indicating that, in fibroblasts, ILK limits RhoA activity.

Targeting of ILK inactivation to fibroblasts was accomplished using a mouse strain carrying the Cre–ER^T fusion transgene under the control of a minimal collagen α2 type I promoter and enhancer driving fibroblast-specific and inducible Cre expression (Fig. 1A) (Sonnylal et al., 2007; Zheng et al., 2002). These mice were crossed to mice in which both Ilk alleles were mutated by the insertion of loxP sites flanking exon 2 (homozygous Ilk^fl/fl) (Grashoff et al., 2003) to produce compound ILK^fl/fl; Cre-positive animals (Fig. 1B). Littermate animals of the genotypes Ilk^fl/+; Cre^pos, Ilk^fl/+; Cre^neg or Ilk^fl/fl; Cre^neg were used as controls. Cre activity was induced in all animals including controls at 12 days of postnatal life by intraperitoneal administration of tamoxifen for five consecutive days as previously described (Zheng et al., 2002; Liu et al., 2009; Sonnylal et al., 2007). Induction of Cre activity was examined in parallel experiments, in which the collagen Cre line was mated to reporter mice harboring a lacZ gene encoding β-galactosidase that was interrupted by a loxP-flanked stop cassette (Soriano, 1999). Cre activity was confirmed by blue staining in the granulation tissue at 7 days after injury, leaving the epidermis unaffected (supplementary material Fig. S1).

Absence of ILK was confirmed in primary fibroblasts isolated from the dermis of tamoxifen-treated mice at 3 weeks of age by immunofluorescence (Col1-ILK; Fig. 1C) and western blot analysis (fl/fl Cre; Fig. 1D). Abundant ILK expression and partial localization in focal adhesion sites was detected in control cells, whereas deleted cells were negative. Very rarely, some deleted fibroblasts exhibited ILK expression (Fig. 1C,D), indicating highly efficient but not complete induction of Cre recombinase activity and ILK deletion in all skin fibroblasts.

By clinical inspection, the skin of Col1-ILK mice appeared inconspicuous before and directly after tamoxifen administration. In addition, histological analysis of the unwounded skin from these mice did not reveal histomorphological differences from controls (supplementary material Fig. S2). We then challenged the animals after tamoxifen administration by creating full-thickness wounds on their backs as previously described (Zweers et al., 2007). Cross-sections taken from the center of wounds at 7 days after injury revealed complete re-epithelialization in all animals. There was a tendency towards acanthosis in control wounds but, overall, the overlying neo-epidermis appeared normal (Fig. 2A). By contrast, granulation tissue formation was severely reduced in wounds of mutant animals (Fig. 2A, lower panel), resulting in greater wound width (Fig. 2A, lower panel, and Fig. 2B) and less granulation tissue (Fig. 2C). Routine histology also indicated markedly reduced cellularity in the granulation tissue, whereas blood vessel numbers appeared unaltered. Moreover, immunostaining revealed normal abundance of macrophages (F4/80 staining, not shown). Myofibroblasts in the granulation tissue were visualized by staining for αSMA, which was detected in the smooth muscle layer of blood vessels in control and mutant wounds (Fig. 2D, arrows). However, myofibroblasts were virtually absent in wounds of Col1-ILK mice, whereas abundant myofibroblasts were observed in control wounds (Fig. 2D, arrowheads). Ki67 staining revealed reduced numbers of proliferating cells in the dermis of mutant mice (Fig. 2E,F). Interestingly, the number of Ki67-positive keratinocytes in the basal layer of mutant epidermis also appeared to be lower, suggesting an altered epithelial–mesenchymal crosstalk.

To determine the mechanisms leading to reduced myofibroblast numbers in mutant wounds and to delineate cell-autonomous effects, we characterized primary fibroblasts that were isolated from the dermis of Col1-ILK and control mice at ~3 weeks after birth, which was around 2–5 days after the last injection with tamoxifen.

To explain the striking absence of myofibroblasts in mutant granulation tissue, we first tested proliferative potential; however, no significant differences were observed in vitro in monolayer cultures (supplementary material Fig. S3A) or in tethered three-dimensional collagen lattices (supplementary material Fig. S3B). We then examined the capacity of ILK-deficient fibroblasts to differentiate to myofibroblasts in vitro, using tethered collagen lattices in which fibroblasts develop forces counterbalancing stress from an ECM environment under tension (Kessler et al., 2001; Tomasek et al., 2002). Fibroblast to myofibroblast conversion was determined by analyzing αSMA expression as the hallmark of myofibroblasts. In comparison with controls, ILK-deficient fibroblasts showed sharply reduced αSMA signals (Fig. 3A).

Because TGFβ1 is a potent inducer of αSMA, we next tested whether reduced formation of myofibroblasts might result from reduced TGFβ1 activity produced by ILK-deficient fibroblasts. Reverse transcription (RT)-PCR and northern blot hybridization revealed similar Tgfb1 mRNA levels (Fig. 3B). We then assessed levels of total (i.e. latent and active, measured following heat activation) and active TGFβ1 protein in the supernatant of fibroblasts cultured in tethered collagen lattices. The assay made use of mink lung epithelial cells, which are engineered to express luciferase driven by the TGFβ1-responsive plasminogen activator inhibitor-1 (PAI-1) promoter (Abe et al., 1994). In contrast to comparable Tgfb1 mRNA levels, amounts of both total and active TGFβ1 protein present in conditioned medium were significantly lower in Ilk-null fibroblasts than in controls (Fig. 3C). The fraction of active TGFβ1 was similar in mutant (19% of total) and control (20% of total) fibroblasts.

Apparently, the levels of fibroblast-generated TGFβ1 were not sufficient to induce the myofibroblast phenotype in vitro. To determine whether ILK-deficient fibroblasts are able to respond to exogenously added TGFβ1 in vitro and which functions might be restored, we assessed selected fibroblast properties and functions in cells grown with 10 ng/ml TGFβ1 in medium supplemented with 1% fetal calf serum (FCS). TGFβ1 signaling was activated in both control and ILK-deficient fibroblasts, as confirmed by phosphorylation of Smad2 (Fig. 4A). Levels of phosphorylated Smad2 were comparable after stimulation with exogenous TGFβ1, indicating that TGFβ receptors are normally expressed and functional on mutant cells. Western blot analysis and quantification of signals demonstrated that supplying exogenous TGFβ1 to ILK-deficient fibroblasts resulted in restoration of αSMA expression to control levels (Fig. 4A). Rescue was confirmed by immunofluorescence analysis using antibodies directed against αSMA. Without TGFβ1, ILK-deficient fibroblasts displayed few αSMA-positive stress fibers, if any (Fig. 4B), whereas addition of TGFβ1 resulted in incorporation of αSMA into clearly visible stress fibers in all ILK-deficient fibroblasts.

The larger wound size of Col1-ILK mice further suggests that mutant fibroblasts are less able than wild-type controls to contract collagen matrices. This defect was indeed observed and was largely rescued by addition of exogenous TGFβ1 (Fig. 4D).

Microscopic inspection of cultured fibroblasts displayed striking differences in cell shape: cells derived from Col1-ILK skin were thin and spread mostly to a needle-like shape (Fig. 5). This morphological difference was observed on integrin β1 substrates, including collagen I (Fig. 5A) and fibronectin (Fig. 5B), and on vitronectin, an integrin β3 substrate (supplementary material Fig. S4). Reduced spreading might be accounted for by reduced surface expression of integrin receptors; however, expression analysis of integrin subunits β1, α2, α5 and αvβ3 by flow cytometry did not reveal significant differences between Ilk-null and control fibroblasts and thus ruled out this explanation (data not shown). Furthermore, no differences in integrin conformation and, hence, ligand affinity and activation status were observed, as indicated by unaltered needle-shaped morphology of mutant fibroblasts upon spreading in medium containing magnesium (supplementary material Fig. S4 and data not shown).

In addition to abnormal cell shape, ILK-deficient fibroblasts failed to organize F-actin filaments in a manner similar to that of controls (Fig. 5A, top panel). Vinculin-positive matrix adhesions in mutant cells appeared shorter and thicker (Fig. 5A, middle panel and insets), and their number was lower than controls (22 per Ilk-null cell versus 91 per control cell; n=100 cells per genotype), indicating disrupted architecture of actin microfilaments and matrix adhesions. In contrast to the restorative effect of exogenous TGFβ1 on αSMA expression, no rescue of cell shape was observed upon treating ILK-deficient fibroblasts with TGFβ1. Abnormal shape and organization of F-actin and matrix adhesions were most pronounced in the early post-adhesion phase, where ILK deficiency clearly resulted in failure of cells to spread (Fig. 5B). Here, phalloidin staining for F-actin revealed a broad, condensed band of cortical actin at the cell periphery and short cytoplasmic cables in fibroblasts lacking ILK. In addition, vinculin was localized diffusely and only occasionally associated with very faint small, rod-like structures at the cell periphery.

The persistence of altered size and shape of ILK-deficient fibroblasts following TGFβ1 treatment suggested the involvement of further mechanisms in cell-shape determination. In particular, the sharply reduced cell size and actin filament arrangement in deficient fibroblasts that was detected after 2 hours of culture raised the question whether the cells might be unable to spread owing to insufficient amounts of active Rac1 and concomitantly elevated RhoA activity. To address this question, we first incubated ILK-deficient and control fibroblasts with Y-27632, which interferes with RhoA signaling by inhibiting ROCK. First, starved cells were seeded for 2 hours in the presence of 5 μM Y-27632 onto fibronectin-coated glass surfaces. Previous results have shown that enhanced spreading by Ilk-null fibroblasts following treatment with inhibitor was most pronounced at 2 hours after addition of the inhibitor and lasted until a maximum of 6 hours (K.B., data not shown). Phase-contrast images show a clear effect with relaxation and enhanced spreading of ILK-deficient fibroblasts (Fig. 6A). Quantification of the cell area at 2 hours after seeding revealed that untreated Ilk-null fibroblasts covered 44% of the area covered by controls (Fig. 6B; P<0.0001; n>224). Rho–ROCK inhibition resulted in expansion of ILK-deficient fibroblasts to 75% relative to untreated controls (P<0.0001; n>224). Parallel cultures were incubated with TGFβ1 (data not shown), which resulted in expansion to 56%: a modest increase in spreading of ILK-deficient fibroblasts in comparison with the area covered by controls, confirming our previous results, indicating that TGFβ1 is unable to normalize cell shape (Fig. 5). To examine the effects of Rho–ROCK inhibition on cells that were already fully spread, we also treated cultures at 24 hours after seeding with Y-27632 for 2 hours (Fig. 6A, right panel). In the absence of inhibitor (−Y), the differences in cell shape between controls and ILK-deficient cells are striking, whereas inhibition of Rho–ROCK signaling by Y-27632 (+Y) induced formation of large lamellipods in some cells.

These results suggest that ILK-deficient fibroblasts exhibit enhanced Rho–ROCK activity. These cells should therefore display higher levels of the ROCK-mediated phosphorylated form of MLC (MLC-P) (Huveneers and Danen, 2009). This was indeed confirmed by western blot analysis, which revealed significantly higher MLC-P levels in ILK-deficient cells compared with control cells (P<0.05; n=4) (Fig. 6C). Treatment with Y-27632 abrogated MLC phosphorylation in both genotypes. As these results provided indirect evidence for enhanced RhoA activity, we used G-linked immunosorbent assay (G-LISA) to directly confirm elevated RhoA activity in ILK-deficient fibroblasts (Fig. 6D).

The failure of ILK-deficient fibroblasts to spread properly could also be due to reduced Rac1 activity. Levels of GTP-loaded Rac1 were therefore assessed in control and ILK-deficient fibroblasts by pull-down assays with serum-starved fibroblasts. Levels of active Rac1 were strongly reduced in ILK-deficient fibroblasts in comparison to controls (Fig. 7A). Because Rac1 activity is also required for the extension and stabilization of lamellipodia at the leading edge of migrating cells (Ridley, 2000), and reduced migration could contribute to the paucity of cells observed in the granulation tissue of Col1-ILK mice, we analyzed migration of growth-arrested cells into scratch wounds by assessing the size of the free space that was not covered by migrating cells. Whereas control cells repopulated the free area within 10 hours, wounds in cultures of ILK-deficient fibroblasts were significantly larger beyond 10 hours (Fig. 7B, white and black bars; P<0.0001; n=6). Suppression of Rho–ROCK signaling by Y-27632 stimulated wound closure in cells of both genotypes, and abolished differences between ILK-deficient and control cells, inferring rescue of migration (Fig. 7B, blue bars).

Having established that ablation of ILK in fibroblasts leads to reduced extracellular TGFβ1 levels, which appear to be responsible for defective conversion of fibroblasts to myofibroblasts, as well as to increased RhoA–ROCK activity, which appears to determine cell shape and migratory capacity, the question arises as to whether these two effects are linked. We addressed this question by assessing levels of total TGFβ1 in supernatants of ILK-deficient and control fibroblasts cultured in stressed lattices in the absence (see also Fig. 3C) and presence of Y-27632. These experiments confirmed reduction of TGFβ1 levels to 46% of control levels in untreated ILK-deficient fibroblasts (Fig. 8). Inhibition of RhoA–ROCK signaling, however, resulted in a significant increase of TGFβ1 to 72% of control levels (P=0.018), which corresponds to the magnitude of rescue in cell area by Y-27632 (Fig. 6B).

We thus conclude that ablation of ILK in fibroblasts induces increased RhoA–ROCK activity and abnormal morphology, which result in decreased extracellular TGFβ1 activity and failure to produce sufficient αSMA for myofibroblast conversion. Concomitantly, Rac1 activity is reduced and cell migration is impaired.

This study addressed the functional relevance of ILK for skin fibroblast functions in vivo and in vitro. Our results suggest that ILK is necessary to provide a structural link for matrix adhesion sites capable of bearing tensile force, to limit RhoA activity and to promote release of TGFβ1, and that it is an essential component of matrix adhesions for the differentiation of fibroblasts to myofibroblasts, their migration into wound sites and re-establishing a normal dermis.

In vivo analysis of ILK function in fibroblasts was achieved by breeding Ilk^fl/fl mice with a strain that carries a tamoxifen-inducible Cre recombinase under the control of a minimal α2(I) procollagen promoter linked to a fibroblast-specific enhancer (Denton et al., 2009; Liu et al., 2009; Sonnylal et al., 2007; Zheng et al., 2002). Efficient deletion of ILK was seen in fibroblasts of Col1-ILK mice, but not in keratinocytes or endothelial cells.

To investigate the contribution of ILK to activation of fibroblasts during tissue repair, we explored granulation tissue formation in excisional wounds. Induced deletion of ILK in fibroblasts in mice did not affect the influx of inflammatory cells, formation of blood vessels or re-epithelialization. However, deletion resulted in perturbed formation of granulation tissue in healing wounds, characterized by virtual absence of myofibroblasts, strongly diminished deposition of dermal matrix and contraction of granulation tissue, and reduced dermal cell density.

To dissect fibroblast-autonomous defects aside from the complex tissue environment, primary fibroblasts were cultured from the dermis of tamoxifen-injected Col1-ILK and control mice. In comparison with controls, ILK-deficient fibroblasts were much smaller and characterized by a needle-like shape, reduced spreading, and disorganized actin cytoskeleton and matrix adhesions on substrates engaging β1 and β3 integrins. Reduced spreading was not due to reduced expression of integrin receptors and probably not to differences in integrin conformation and, hence, in ligand affinity and activation status (Banno and Ginsberg, 2008; Hughes and Pfaff, 1998). Normal surface expression levels of several integrin subunits including β1 integrin were also reported for keratinocytes, chondrocytes and vascular smooth muscle cells lacking ILK (Grashoff et al., 2003; Lorenz et al., 2007; Montanez et al., 2009).

The low cell density observed in wounds of Col1-ILK mice prompted the analysis of proliferative potential. Reduced numbers of Ki67-positive cells were detected in mutant wounds, indicating reduced proliferation of mutant cells. This is in contrast to our in vitro findings, which show that proliferation of Ilk-null fibroblasts in two- or three-dimensional cultures did not differ significantly from controls. This discrepancy may be explained by the complex wound environment, in which mutant fibroblasts might not be able to respond adequately to the different triggers. Impaired migration of ILK-deficient fibroblasts into the wound area might further add to an overall lower cell density within the wound.

Virtually all cell types rendered ILK-deficient, ranging from Caenorhabditis elegans (Norman et al., 2007), zebrafish (Postel et al., 2008), Drosophila (Zervas et al., 2001) to mice (Gheyara et al., 2007; Hannigan et al., 2007; Lorenz et al., 2007; Nakrieko et al., 2008; Sakai et al., 2003; this report), exhibit disruption of actin cytoarchitecture and matrix adhesions. In fibroblasts, these structures are prerequisites for fulfilling dynamic functions (Bershadsky et al., 2006; Goffin et al., 2006) and, accordingly, spreading, migration and matrix contraction by ILK-deficient fibroblasts were strongly impaired. In contrast to other cell types and specific to fibroblasts is the observed differentiation defect, which results in impaired formation of myofibroblasts. We conclude that generation of this specialized ECM-synthesizing, contractile, αSMA-positive fibroblast is precluded because matrix adhesions without ILK assemble into structurally abnormal complexes that fail to bear or transmit the required mechanical tension. The structural defect in ILK-deficient fibroblasts could be the absence of the link between ILK and PINCH, parvin and/or α-PIX similarly to that in contractile smooth muscle cells, in which ILK regulates neuronal Wiskott-Aldrich syndrome protein (N-WASP)-mediated actin polymerization, cell contraction and development of mechanical tension (Zhang et al., 2007). Kindlin-2 is another protein that binds directly to ILK in matrix adhesions of fibroblasts and also to integrin β1 tails, and cooperates with talin in integrin activation (Harburger et al., 2009; Larjava et al., 2008; Montanez et al., 2008). Kindlin-2 expression has been shown to be retained in ILK-deficient fibroblasts (Montanez et al., 2008), where it might provide some functional compensation for the loss of ILK.

Specific defects in mechanotransduction have also been reported in ILK-deficient fibroblast-like cells, which were not derived from the dermis of mouse skin, but from embryos (Maier et al., 2008). Upon mechanical loading in vitro, these cells failed to exhibit responses typical for mechanically stressed fibroblasts. By contrast, phosphorylation of ERK1 and/or ERK2, focal adhesion kinase and protein kinase B (PKB, also known as Akt) in response to stress was identical in ILK-deficient and control cells.

Besides mechanical tension, myofibroblast formation strictly requires TGFβ1, which is pivotal for fibroblast function in inflammation, scarring and fibrosis (Denton et al., 2009; Hinz, 2007; Tomasek et al., 2002). Synthesis of TGFβ1 itself is positively regulated by mechanical stress in fibroblasts at the transcriptional level (Kessler et al., 2001). Moreover, TGFβ1 can be activated from extracellular stores by mechanical deformation of LAP in an integrin-dependent mechanism (Wipff et al., 2007). Our results suggest that gene expression and the percentage of latent TGFβ1 conversion to its active form are not impaired in ILK-deficient fibroblasts in collagen lattices under stress. However, levels of extracellular active and total TGFβ1 in conditioned medium appeared to be coordinately regulated and were both strongly reduced in ILK-deficient fibroblasts. This strict correlation between the amount of latent TGFβ1 available to the cells and the TGFβ1 activity generated has been observed previously, e.g. in fibroblast–keratinocyte co-cultures (Shephard et al., 2004), and is apparently not due to auto-induction of TGFβ synthesis. Thus, our data indicate that Ilk-null fibroblasts have a defect in TGFβ homeostasis, i.e. the cells are impaired in their capacity to maintain extracellular TGFβ1 levels. Regarding potential mechanisms explaining this defect, reduced transcription and RNA stability can be ruled out, as Tgfb1 RNA steady-state levels are similar in control and Ilk-null fibroblasts, pointing to defects in post-transcriptional events. These could be related to decreased protein synthesis, altered intracellular modification, defective processing of pro-TGFβ (Yang et al., 2010) or altered assembly of the large latent complex, all of these potentially resulting in decreased latent TGFβ1 secretion and activation (Miyazono et al., 1991; Saharinen et al., 1996). Decreased secretion of factors including TGFβ1 has been reported for other cell types that lack ILK. TGFβ1 secretion was impaired in mesangial cells transfected with a kinase-dead mutant of ILK (Ortega-Velazquez et al., 2004). Inhibition of ILK by knockdown (Lin et al., 2007) or by small molecule inhibitors (Edwards et al., 2008) also resulted in selected suppression of secreted vascular endothelial growth factor and urokinase plasminogen activator levels, but not those of PAI-1, indicating that ILK is involved in specific intracellular protein processing events and/or secretory pathways. Alternatively, proteolytic mobilization of latent TGFβ1 from extracellular growth factor deposits, on the cell surface or in the ECM, might be impaired, leading to reduced levels of TGFβ1 precursor available to the cells for activation.

TGFβ receptors are functional in mutant fibroblasts, as was demonstrated by normal Smad signaling following addition of exogenous TGFβ1. This treatment was sufficient to rescue production of αSMA and its incorporation into stress fibers, confirming the strict dependence of this differentiation process on TGFβ1, and to largely rescue collagen lattice contraction. These data highlight the importance of TGFβ1 activity in the immediate vicinity of fibroblasts for myofibroblast formation. In healing wounds, TGFβ1 is thought to be supplied mainly by platelets and macrophages, particularly at early stages (Eming et al., 2007). Other cell types, such as fibroblasts, keratinocytes and vascular cells, probably contribute to latent TGFβ secretion and/or activation (Brunner and Blakytny, 2004), resulting in a second peak of TGFβ activity upon wound closure (Yang et al., 1999). Because the influx of inflammatory cells, as well as angiogenesis, were not altered in our system, it appears that the fibroblasts themselves are responsible for mobilizing and activating sufficiently high amounts of TGFβ1 in the immediate vicinity of fibroblasts and that this is crucial to ensure myofibroblast differentiation. This is further supported by the observation that keratinocyte proliferation is reduced rather than increased. Since TGFβ1 is a potent negative regulator of epithelial growth (Massague, 2000; Shipley et al., 1986), this observation suggests that reduced release of TGFβ1 by mutant fibroblasts is a local phenomenon that is restricted to the immediate fibroblast environment, and does not result in generally lower TGFβ1 concentrations in the healing wound.

Addition of exogenous TGFβ1 clearly did not change the small size, lack of spreading and overall fibroblast morphology that was caused by inactivation of ILK. Concomitant with low Rac1 activity, we found increased RhoA–ROCK activity and enhanced phosphorylation of MLC in ILK-deficient fibroblasts, which was inhibited by Y-27632. Suppression of RhoA–ROCK signaling enabled almost normal cell spreading and migration. The inhibitory effects by Y-27632 on levels of MLC-P, cell morphology and migration, as well as TGFβ levels, were extensive, whereas an increase in RhoA activity in ILK-deficient fibroblasts was surprisingly less pronounced. An explanation for this might be that Y-27632 also modulates the activity of other kinases beyond ROCK, such as PKCε (Ishizaki et al., 2000; Uehata et al., 1997). Although the affinity of Y-27632 is markedly higher for Rho-associated kinases, other kinases could potentially also contribute to the observed defects in ILK-deficient fibroblasts.

The observation of delayed collagen lattice contraction by ILK-deficient cells is in agreement with our previous work, which reported delayed collagen lattice reorganization and contraction by fibroblasts expressing constitutively active RhoA (Zhang et al., 2006b), and can be explained by the concept that lattice reorganization by fibroblasts is mediated by cell migration and traction (Harris et al., 1981).

We conclude that, in fibroblasts, ILK functions as a negative regulator of RhoA activity (Montanez et al., 2009). The exact mechanism by which ILK controls RhoA signaling remains elusive. Interestingly, a recent report indicates that the ILK–α-parvin complex is involved in limiting RhoA activity. Thus, ablation of α-parvin in vascular smooth muscle cells resulted in a similar phenotype to the one described here, with similarly elevated RhoA levels and spreading defects. The phenotype is more severe in fibroblasts, probably because the ILK–α-parvin complex cannot assemble (Legate et al., 2006), whereas in the mural cells used by Montanez and colleagues, an independent ILK–β-parvin complex forms and compensates, to some extent, for the absence of α-parvin (Montanez et al., 2009). Mechanistically, high RhoA activity might be achieved by lack of direct inhibitory action of Rac1 on RhoA, or, indirectly, by low amounts or activity of Rho GTPase activating proteins, which normally limit amounts of the active, GTP-bound form of Rho. Increased RhoA–ROCK activity was further reported in cells with defective spreading and migration following transfection with EphA1, a tyrosine kinase receptor for ephrin-A. Interestingly, ILK was downregulated in these cells (Yamazaki et al., 2009).

Our current results do not allow to us conclude whether ablation of ILK in fibroblasts first results in increased RhoA–ROCK activity, which then stimulates decreased Rac1 activity, or vice versa. However, they show that TGFβ1 production by fibroblasts is under the control of Rho activity, possibly by regulating post-transcriptional events involved in processing, assembling, secreting or mobilizing the large latent TGFβ complex. In summary, this study reveals the central function of ILK in the generation of myofibroblasts by two mechanisms. First, it serves as a structural adaptor, orchestrating the proper topology of integrin-containing matrix adhesions, which are required for efficient mechanotransduction. Second, ILK controls fibroblast morphology and migration by limiting Rho–ROCK signaling, which, in turn, regulates TGFβ1 production required for myofibroblast conversion and mechanical tension. Impairment of these functions explains the insufficient myofibroblast differentiation and migration into granulation tissue, which compromises dermal wound healing in mice with fibroblast-restricted deletion of ILK, but might be beneficial in protection from fibrosis. Although the effects are most pronounced in fibroblasts, these mechanisms are not restricted to these cells and, in fact, appear to operate in several cell types (Montanez et al., 2009) and might thus be of general biological significance.

Mice with a floxed exon 2 containing the translation start ATG (Grashoff et al., 2003; Sakai et al., 2003) were bred to a mouse line that expresses inducible Cre recombinase under the control of a fibroblast-specific regulatory fragment of the pro-α2(I) collagen gene (Col1a2) (Zheng et al., 2002). Genotyping was performed using PCR for the presence of the floxed (370 bp) or wild-type (296 bp) Ilk allele (forward primer 5′-GTCTTGCAAACCCGTCTCTGCG-3′; reverse primer 5′-CAGAGGTGTCAGTGCTGGGATG-3′) and for Cre (forward primer 5′-GACGGAAATCCATCGCTCGA-3′; reverse primer 5′-GACATGTTCAGGGATCGCCA-3′). Cre activity was initiated by intraperitoneal injection of 1 mg of tamoxifen (20 mg/ml stock in sunflower seed oil; Sigma) per mouse per day on five consecutive days into 12-day-old mice. Animals were housed in specific pathogen-free facilities. All animal protocols were approved by the local veterinary authorities.

Female animals that carried the Cre gene and were homozygous for the Ilk^fl allele (Col1-ILK) were used as experimental animals at ~3 weeks of age (2–5 days after the last tamoxifen injection). Females that were either Ilk^fl/+; Cre-positive or Cre-negative were used as controls. All animals were derived as offspring from matings between Ilk^fl/+; Cre-positive and Ilk^fl/+; Cre-negative animals on a mixed background. On the shaved backs, one full-thickness wound, comprising the skin and subcutaneous panniculus carnosus muscle, of 4 mm in diameter was inflicted on each side of the scapular region using biopsy punches under isoflurane anesthesia and left uncovered to heal.

Wounds were bisected and either frozen unfixed in optimal cutting temperature (OCT) compound (Sakura) or fixed for 3 hours in 4% paraformaldehyde solution before paraffin embedding. Cryosections of 6 μm were mounted on polylysine-coated slides and fixed for 30 minutes in ice-cold methanol. For general histology, paraffin-embedded sections were stained with hematoxylin and eosin (H&E) or Masson's trichrome following standard procedures. Cell proliferation was assessed by immunohistochemistry using antibodies against Ki67 (TEC-3, Dako), followed by biotin-conjugated secondary antibodies. For immunofluorescence, sections were blocked with the appropriate normal serum (Dako) before application of primary antibodies. Cells were permeabilized (0.2% Triton X-100 in PBS, 10 minutes at room temperature) before blocking. Primary antibodies were as follows: αSMA (Cy3-conjugated clone 1A4, Sigma), ILK (I-1907, Sigma), phalloidin–Alexa-Fluor-594 (Molecular Probes, Invitrogen) and vinculin (MAB 3574, Chemicon, Millipore). Secondary antibodies were coupled to Alexa Fluor 488 or Alexa Fluor 594 (Molecular Probes, Invitrogen). Nuclei were counterstained either with propidium iodide or DAPI. Photomicrographs were taken using either a DM 4000B microscope (Leica) or a Nikon Eclipse 800E fluorescence microscope equipped with a digital camera (DXM 1200F; Nikon). Image analysis was performed using Lucia software (Laboratory Imaging) or DISKUS (Hilgers).

Fibroblasts were cultured from the dermis of 3-week-old animals as previously described (Zweers et al., 2007) after tamoxifen induction and genotyping, and were used between passages 1 and 6. Briefly, trunk skin was minced and digested with 400 units/ml of collagenase I (Cell Systems) for 2 hours at 37°C. The homogenate was cleared by passing through a cell sieve (Falcon), and cells were cultured in DMEM (Invitrogen) containing 10% FCS (Thermo Scientific), 50 μg/ml Na-ascorbate (Sigma), 2 mM glutamine and antibiotics (Seromed-Biochrom) at 37°C in 5% CO₂ on tissue-culture plastic (Becton Dickinson Labware). Because prolonged culture of incompletely deleted fibroblast populations can result in overgrowth of undeleted cells, genotypes were routinely confirmed in all experiments. For immunofluorescence, cells were seeded on Permanox chamberslides or on chambers with glass bottoms (Nunc, Thermo Scientific) that were coated with bovine collagen I (30 μg/ml in PBS; IBFB, Curacyte) or human fibronectin (10 μg/ml; Sigma). TGFβ1 (recombinant human, R&D Systems) was added to serum-deprived cultures at a concentration of 10 ng/ml in DMEM with 1% FCS directly at culture onset for 24 hours. ROCK inhibitor Y-27632 (Calbiochem) was added at 5 μM as specified.

Fibroblasts were starved overnight in DMEM with 1% FCS, ascorbate, antibiotics and glutamine, and then seeded in the same medium at 7×10⁴ cells per gel (500 μl) into 24-well plates with 0.3 mg/ml of acid-extracted collagen I from newborn calf skin (IBFB, Curacyte) as previously described (Kessler et al., 2001; Zhang et al., 2006a). Lattices were detached manually to facilitate contraction. Tethered lattices were cast with 2×10⁵ fibroblasts per lattice (6 ml), as described above, into 6 cm dishes. Braided nylon rings inserted into the dishes kept the lattices from shrinking (Kessler et al., 2001). For western blot analysis, fibroblasts were liberated from the lattices by incubation with 400 U/ml collagenase I (Cell Systems) in serum-free DMEM for 15 minutes at 37°C.

Fibroblasts were growth-arrested by mitomycin C treatment (Sigma; 4 μg/ml for 2 hours at 37°C) and seeded at high density (50,000 cells/cm²) in DMEM with 1% FCS. Scratch wounds of ~1 mm width were introduced with a disposable pipette tip. Cell migration into the denuded area was monitored for 10 hours in the absence or presence of 5 μM Y-27632 in an incubator chamber attached to an IX81 microscope (Olympus) at 37°C, 5% CO₂ and 60% humidity. Frames were taken every 30 minutes using an OBS CCD FV2T camera, and the extent of remaining free area was quantified using OBS Cell R software (Olympus).

TGFβ1 levels were assessed in 24-hour conditioned medium of tethered collagen lattice cultures that were cast in DMEM with 1% FCS with or without 5 μM Y-27632. TGFβ activity was determined in the plasminogen activator inhibitor-1 (PAI-1)/luciferase (PAI/L) assay (Abe et al., 1994) using mink lung epithelial cells stably transfected with a luciferase reporter gene under the control of a TGFBβ-specific, truncated PAI-1 promoter (a gift from Daniel B. Rifkin, New York University Medical Center, New York, NY). Samples were assayed at two dilutions to fit into the linear range of the assay (5–250 pg/ml). Total TGFβ activity was determined following heat activation of the samples for 10 minutes at 80°C. Following overnight incubation of the reporter cells with the samples or with recombinant human TGFβ1 standards, cells were lysed and luciferase activity determined using an MLX Microtiter Plate Luminometer (Dynex Technologies).

Rac activity was assayed with a biotinylated peptide corresponding to the CRIB motif of the Rac1 effector Pak, essentially as described previously (Price et al., 2003). Confluent fibroblast cultures were starved overnight in medium without FCS and activated by addition of 10% FCS for 45 minutes. Cells were washed with PBS and scraped off the plates into 700 μl lysis buffer (50 mM Tris-HCl pH 7.4, 100 mM NaCl, 10 mM MgCl₂, 1% NP-40 and protease inhibitor mix; Sigma). Three μl of the CRIB peptide (2 mg/ml; Biotrend) were added, and lysates were rotated for 45 minutes at 4°C. After clearing by centrifugation, an aliquot of total lysates was removed and remaining lysates were rotated another 45 minutes at 4°C. The latter were then incubated with 30 μl streptavidin-agarose beads (Novagen, Merck) for 30 minutes at 4°C. The beads were washed, and bead samples and equal amounts of total lysate were separated by SDS-PAGE. Rac was detected by western blotting using an anti-Rac1 antibody (clone 23A4; Upstate Biotechnology, Millipore). RhoA activity was assessed using G-LISA (#BK124, Cytoskeleton) according to the manufacturer's protocol. Serum-starved cells were reseeded at 400,000 cells per 6 cm culture dish for 6 hours, stimulated for 5 minutes with FCS and lysates were snap-frozen in liquid nitrogen prior to applying to G-LISA assay.

Northern blot hybridization was carried out as previously described (Kessler et al., 2001) using ³²P-labeled probes. The TGFβ1 cDNA probe was generated by reverse transcription of total wild-type mouse RNA (Fermentas) and PCR amplification of a 300 bp fragment of murine TGFβ1 (forward primer 5′-CAACGCCATCTATGAGAAAACC-3′; reverse primer 5′-AAGCCCTGTATTCCGTCTCC-3′). Hybridization with a probe detecting 18S rRNA was used to control for equal loading as previously described (Kessler et al., 2001). TGFβ1 transcript levels were also assessed by semi-quantitative RT-PCR using forward 5′-CAACGCCATCTATGAGAAAACC-3′ and reverse 5′-AAGCCCTGTATTCCGTCTCC-3′ primers to amplify a 300 bp product of TGFβ1 cDNA (Fermentas) and forward 5′-AATGTGCAGCCCATTCGCTG-3′ and reverse 5′-CTTCCGTCCTTACAAAACGG-3′ primers for the product (324 bp) of S26 cDNA, which was used for normalization.

Proteins for western blot analyses were extracted in RIPA buffer (150 mM NaCl, 50 mM Tris-HCl, 1% NP-40, 0.1% SDS, 0.5% Na-deoxycholate, 5 mM EDTA, 1% Triton X-100, pH 8), supplemented with protease and phosphatase inhibitor cocktails (Sigma), separated by SDS-PAGE and transferred to Hybond C extra membranes (Amersham Biosciences). Fibroblasts in tethered collagen lattices were first liberated by digestion with collagenase (400 U/ml CLS1, for 15 minutes at 37°C in serum-free DMEM; Cell Systems) before lysis in RIPA buffer. Following blocking (5% milk powder in TBS-T, pH 7.6), membranes were incubated with primary antibodies in blocking solution. Signals were detected by chemiluminescence using Western Lightning (Perkin Elmer) or SuperSignal West Femto (Thermo Scientific). Primary antibodies used were directed against αSMA (clone 1A4, Sigma), ILK (I-1907, Sigma), phosphorylated Smad2 (recognizing Ser465 and Ser467; Cell Signaling/New England Biolabs #3101), MLC (clone MY-21, Sigma), phosphorylated MLC (#3674, Cell Signaling) and actin (MP Biomedicals). Secondary HRP-conjugated antibodies were purchased from Dako. Identical gels run in parallel and stained with Coomassie Blue were used as loading control. Signals were quantified by densitometry using ImageJ (http://rsbweb.nih.gov/ij).

Results are presented as means ± s.d. Statistical analysis was performed using unpaired Student's t-tests, and P<0.05 was considered statistically significant. Calculations were performed using GraphPad Prism (GraphPad Software)

We thank Roswitha Nischt, Paola Zigrino, Zhigang Zhang, Monique Aumailley, Sabine Eming and Carien Niessen (Cologne) for insightful discussion; Gabi Scherr, Nicole Pohl, Ute Hillebrand (Cologne) and Tamara Berger (Münster) for excellent technical assistance; Carien Niessen for the generous gift of PAK-CRIB peptide; and Kristina Behrendt (Cologne) for help with the Rac pull-down. This work was supported by Deutsche Forschungsgemeinschaft through EC140/5-1 and SFB 829 (to B.E. and T.K.).