Expression of integrin β1 by fibroblasts is required for tissue repair in vivo

During normal healing of skin, connective tissue is repaired exclusively through the action of fibroblasts, which migrate into the wound site where they synthesize and remodel a new extracellular matrix (ECM) (Martin, 1997; Eckes et al., 1999; Gurtner et al., 2008). The ability of cells to migrate, attach to and remodel ECM is mediated by specialized cell surface structures termed focal adhesions (FAs), which mediate adhesion between the ECM and the actin cytoskeleton through integrin cell surface receptors (Burridge and Chrzanowska-Wodnicka, 1996; Zaidel-Bar et al., 2007). The fibroblasts that generate the adhesive and biomechanical forces required for wound closure are termed myofibroblasts, because they express highly contractile proteins including α-smooth muscle actin (α-SMA) (Dugina et al., 2001; Chen et al., 2005; Hinz et al., 2007). Although myofibroblasts normally disappear from the healed wound, the persistence of the myofibroblast appears to be a major contributor to the phenotype of excessive scarring and fibrotic disorders (Dugina et al., 2001; Chen et al., 2005; Hinz et al., 2007; Varga and Abraham, 2007). Controlling myofibroblast action is therefore essential to control normal tissue repair and to develop novel, effective and rational anti-fibrotic therapies.

Several growth factors and cytokines contribute to wound closure in vivo (Werner and Grose, 2003; Barrientos et al., 2008); however, it is now increasingly clear that proteins of the focal adhesions might act as a common integration point for signals emanating from growth factors and their receptors (Burridge and Chrzanowska-Wodnicka, 1996; Zamir and Geiger, 2001) and from interactions with the external environment (Zaidel-Bar et al., 2007; Geiger et al., 2009). As an example, fibroblasts lacking focal adhesion kinase (FAK) spread poorly, cannot migrate, and cannot properly respond to TGFβ (Ilic et al., 1995; Liu et al., 2007). Recently, we have shown that fibroblast expression of Rac1, a small GTPase that is expressed ubiquitously and is recruited to FAs by paxillin (Burridge and Wennerberg, 2004; Ishibe et al., 2004), is required for fibrogenesis and cutaneous tissue repair in vivo (Liu et al., 2008; Liu et al., 2009a) and for induction of myofibroblast formation by endothelin-1 (Shi-wen et al., 2004). Moreover, we have shown that activation of cell adhesion is sufficient to cause increased expression of mRNAs encoding fibrogenic proteins (Kennedy et al., 2008). Collectively, these results suggest that adhesion and adhesive signaling has a key role in tissue repair and fibrogenic responses in vivo.

Given their role in cell adhesion and adhesive signaling, integrin expression by fibroblasts must be important for tissue repair in vitro, although the contribution of individual integrins to this process is still being elucidated (Hynes, 2002; Schultz and Wysocki, 2009). Integrin β1 is a common mediator of fibroblast attachment to collagen type I and fibronectin (Lafrenie and Yamada, 1996; Gabbiani, 2003), suggesting that expression of integrin β1 by fibroblasts has a key role in cutaneous tissue repair. However, this hypothesis has not yet been tested. As integrin-β1-knockout mice die in utero (Stephens et al., 1995), conditional-knockout mouse models are necessary to address this question. Mice harboring the integrin β1 allele flanked by loxP sites have been generated, and these mice have been used successfully to delete integrin β1 in vivo using mice expressing Cre recombinase under the control of a tissue-specific promoter (Raghavan et al., 2000; Piwko-Czuchra et al., 2009; Bauer et al., 2009). Recently, we have shown that mice containing a fibroblast-specific deletion of integrin β1 show resistance to bleomycin-induced skin fibrosis (Liu et al., 2009b). In this report, we study the influence of deletion of integrin β1 in adult fibroblasts on cutaneous wound healing and the underlying cell and molecular processes involved. Our data provide new and valuable insights into the contribution of integrin β1 to fibroblast biology.

To test whether integrin-β1-deficient mice showed impaired cutaneous wound closure, we subjected 8-week-old mice homozygous for deletions in the integrin β1 gene (Itgb1), or wild-type littermate counterparts, to dermal punch wounding. Loss of β1 integrin was shown by real-time PCR and western blot analysis of fibroblasts cultured from control (C/C) and integrin-β1-deficient (K/K) animals (Fig. 1A,B). Compared with wounded control animals, integrin-β1-deficient animals showed a significantly reduced rate of wound closure (Fig. 1C). Studies using wild-type mice treated with or without tamoxifen revealed that tamoxifen alone did not affect wound closure (not shown). Integrin-β1-deficient animals were examined 7 days after wounding. They displayed reduced collagen production and less granulation tissue (Fig. 2), as well as myofibroblasts with decreased α-SMA and PCNA expression at the site of injury (Fig. 3) compared with control animals.

To study the molecular mechanism(s) of failed wound repair, we cultured primary dermal fibroblasts from explants of animals deleted for integrin β1 or not. Using a scratch wound assay of in vitro wound repair, we showed that integrin-β1-deficient cells migrated more slowly (Fig. 4A). Single-cell tracking confirmed that persistence of migration in integrin-β1-deficient fibroblasts on collagen I and on tissue culture plastic was significantly diminished (Fig. 4B). Integrin-β1-deficient cells also showed reduced proliferation (Fig. 5A) and an impaired ability to adhere to a fibronectin substrate (Fig. 5B) than observed with control cells. Cell spreading was also monitored microscopically after adhesion, using antibodies against Rhodamine-phalloidin and vinculin. Loss of integrin β1 also resulted in abnormal spreading on fibronectin, as revealed with anti-vinculin antibody (green) and Rhodamine-phalloidin (red) staining to detect actin (Fig. 5C).

Myofibroblasts are the major cell type responsible for ECM remodeling during wound healing (Dugina et al., 2001; Hinz et al., 2007). We showed that integrin-β1-deficient fibroblasts contained reduced numbers of α-SMA-containing stress fibers (Fig. 6A) compared with control cells, suggesting that integrin-β1-deficient cells had impaired myofibroblast differentiation. Consistent with this observation, integrin-β1-deficient fibroblasts also had reduced α-SMA protein expression (Fig. 6B). Moreover, loss of integrin β1 resulted in cells that displayed reduced expression of mRNAs encoding profibrotic proteins, including α-SMA and type I collagen (Fig. 6C). Finally, integrin-β1-deficient fibroblasts were less able to migrate and thus to generate contractile forces to contract free-floating and tethered collagen gel matrices (Fig. 7A,B). These data suggest that integrin β1 deficiency results in a dramatic failure of fibroblast attachment to collagen and differentiation into myofibroblasts.

It has been proposed that integrins (in particular integrin αv) mediate the activation of latent TGFβ during myofibroblast contraction (Wipff et al., 2007). To test whether loss of integrin β1 resulted in impaired activation of latent TGFβ, we co-cultured in collagen gel matrices, fibroblasts and mink lung reporter cells stably transfected with plasminogen activator inhibitor-1 promoter-luciferase construct, which responds to TGFβ (Abe et al., 1994; Wipff et al., 2007). As anticipated, luciferase activity was elevated in gels populated with wild-type cells, but not in gels populated with integrin-β1-deficient cells (Fig. 8). The activity observed in the presence of wild-type cells was significantly inhibited by the addition of the small molecule inhibitor of the TGFβ type I receptor (ALK5) or the neutralizing anti-TGFβ antibody 1D11, indicating the involvement of TGFβ in this process (Fig. 8A,B). These data strongly suggest that integrin-β1-knockout fibroblasts possess decreased TGFβ activation in bio-mechanically loaded gels, which is characterized by the formation of myofibroblasts. To test the mechanism underlying integrin-β1-mediated TGFβ activation, assays were performed in the presence of DMSO and the myosin ATPase contraction inhibitor blebbistatin. Matrix remodeling and ECM contraction in floating gel matrices involves actin or myosin contraction (Daniels et al., 2003; Mirastschijski et al., 2004; Eastwood et al., 1998); thus blebbistatin blocked collagen gel contraction mediated by wild-type integrin β1 cells in floating collagen gel matrices (Fig. 8C). Blebbistatin reduced TGFβ activation in gels embedded with wild-type cells (Fig. 8C). Thus TGFβ activation in floating collagen gel matrices depends on both myosin ATPase and contraction. Addition of TGFβ1 rescued the contractile and TGFβ activation defects of integrin-β1-knockout cells (Fig. 8C).

Western blot analyses of wild-type (C/C) and integrin-β1-deficient (K/K) cells showed no difference in levels of TGFβ type I or TGFβ type II; however, integrin-β1-deficient cells showed elevated integrin β3 expression (Fig. 9), an integrin that has previously been associated with decreased rates of tissue repair (Reynolds et al., 2005).

Based on these results, we investigated whether phenotype of integrin-β1-deficient mice could be rescued by addition of exogenous active TGFβ1 to mice subjected to the dermal punch model of tissue repair. By day 3 after wounding, addition of TGFβ1 resulted in increased wound closure in both wild-type and integrin-β1-deficient mice (Fig. 10A). By day 5 and day 7 after wounding, TGFβ1 increased wound closure in integrin-β1-deficient mice to the level observed in wild-type mice (Fig. 10A). In addition, application of TGFβ1 to day 7 wounds of integrin-β1-deficient mice resulted in levels of granulation tissue, collagen, and α-SMA production that were similar to those observed with the control-injected wild-type mice (Fig. 10B–D). Taken together, these data indicate that the addition of exogenous active TGFβ effectively rescued the wound-healing defects of integrin-β1-deficient mice. These data strongly suggest that expression of integrin β1 by fibroblasts is required for efficient wound closure, specifically proper granulation tissue formation and dermal repair, via a TGFβ-dependent mechanism.

Integrin β1 is essential for embryonic development and the maintenance of several tissues (Brakebusch and Fässler, 2005; Fässler and Meyer, 1995). Integrin-β1-knockout mice die in utero (Stephens et al., 1995), indicating an essential role for this integrin in development. Cell-type-specific deletions of the integrin β1 gene has revealed a key role for integrin β1 in supporting stem and progenitor cell properties, such as adhesion to their niche and the appropriate orientation of the mitotic spindle to control the symmetry of cell division (Lechler and Fuchs, 2005; Kuang et al., 2007). Mutant mice lacking integrin β1 in skin epidermis possessed severe blistering as a result of impaired attachment of basal keratinocytes to the basement membrane, impaired keratinocyte proliferation, dermal fibrosis and severely delayed wound healing (Raghavan et al., 2000; Brakebush et al., 2000; Grose et al., 2002). However, the role of integrin β1 in fibroblasts and myofibroblast-like cells in vivo is largely unknown, as is the precise cell and molecular mechanism underlying integrin β1 action via fibroblasts in tissue repair.

Integrin β1 is a key mediator of fibroblast adhesion to type I collagen and fibronectin (Lafrenie and Yamada, 1996; Gabbiani, 2003). In this study, we show that loss of integrin β1, specifically in adult fibroblasts, results in impaired cutaneous tissue repair in vivo. Integrin-β1-deficient mice showed reduced rates of wound closure, and decreased myofibroblast production and collagen deposition. Consistent with these results, cultured fibroblasts derived from integrin-β1-deficient mice possessed decreased levels of α-SMA mRNA and protein expression. Consistent with these data, integrin-β1-deficient fibroblasts exhibited reduced α-SMA stress fiber formation and reduced expression of mRNA encoding type I collagen and connective tissue growth factor (CTGF). Moreover, integrin β1 was required for optimal cell migration, adhesion and ECM contraction. These results are in sharp contrast to the conventional knock-out mice lacking integrin β3, which show enhanced wound healing that is complete several days earlier than in wild-type mice, and enhanced dermal fibroblast infiltration into wound sites (Reynolds et al., 2005). Interestingly, we found that loss of integrin β1 resulted in an increase in expression of integrin β3, providing a possible mechanistic basis for our study. It is also interesting to note that wound healing in the integrin-β2-knockout mouse was similar to that in our fibroblast-specific knockout of β1 integrin (Peters et al., 2005). Here, the mechanisms underlying the severely impaired wound healing were associated with a severe reduction of neutrophils and macrophages into the wounds, causing a lack of secreted TGFβ1. As our mice are only deficient in integrin β1 in fibroblasts, the molecular mechanism underlying impaired healing is likely to be different. Indeed, it was previously suggested that mechanical contraction of ECM by fibroblasts activates TGFβ in a fashion requiring integrin αV and an intact cytoskeleton (Wipff et al., 2007). To determine whether such a mechanism might be responsible for the phenotype of integrin-β1-knockout fibroblasts, we showed that wild-type fibroblasts, but not integrin-β1-deficient fibroblasts, embedded within a collagen gel were able to support high levels of activity of a TGFβ-responsive promoter in a fashion that depended on myosin-mediated contraction. TGFβ rescued the phenotype of integrin-β1-deficient mice. These results collectively suggest that expression of integrin-β1 by fibroblasts is essential for wound healing, and that the function of integrin β1 occurs partly through a TGFβ-dependent mechanism probably through a defect in activation of latent TGFβ. Our results demonstrate for the first time that integrin β1 expression by fibroblasts is involved with this process and is required for formation of granulation tissue in skin.

Fibrosis is characterized by the persistence of myofibroblasts within scars (Dugina et al., 2001; Chen et al., 2005; Gabbiani, 2003). Although myofibroblasts can form through the presence of extracellular signaling molecules including TGFβ, ET-1 and CCN2, it is now increasingly appreciated that adhesive signaling also contributes to their formation (Werner and Grose, 2003; Leask, 2006; Shi-wen et al., 2006a,b; Shi-wen et al., 2007). Our data showing that integrin β1, a protein that mediates cell attachment of ECM, was required for tissue repair in vivo is consistent with this latter notion. Taken together with observations that the persistent fibrotic phenotype of lung scleroderma fibroblasts is at least partially mediated by integrin β1 (Shi-wen et al., 2007), which is upregulated in scleroderma fibroblasts (Chen et al., 2005; Chen et al., 2008), our results strongly suggest that the targeting of adhesion signaling is a novel, appropriate strategy for anti-fibrotic drug intervention. In summary, the observations presented here indicate that integrin β1 is essential for proper granulation tissue formation and thus for wound closure and might have future implications for the treatment of non-healing ulcers and fibrosis.

Integrin β1 conditional-knockout mice were generated as described (Liu et al., 2009b). Briefly, mice possessing a tamoxifen-dependent Cre recombinase under the control of a fibroblast-specific regulatory sequence from the proα2(I) collagen gene (Zheng et al., 2002) were crossed with mice homozygous for a conditional Itgb1 (integrin β1) allele (Raghavan et al., 2000) (Jackson Laboratories, Bar Harbor, ME) to generate Cre and integrin β1 heterozygous mice, which were mated to generate Cre integrin β1 homozygous mice. Animals used for experiments were genotyped by polymerase chain reaction (Raghavan et al., 2000; Zheng et al., 2002). To delete integrin β1, 3-week-old mice were given intraperitoneal injections of tamoxifen suspension (0.1 ml of 10 mg/ml 4-hydroxytamoxifen, Sigma, St Louis, MO) over 10 days. Deletion of integrin β1 was tested by PCR genotyping. All animal protocols were approved by the appropriate regulatory authority.

Littermate mice homozygous for the loxP Itgb1 allele and heterozygous for type I Col1A2-Cre were treated with tamoxifen (‘knockout integrin β1’, K/K) or corn oil (‘conditional integrin β1’, C/C). Three weeks after cessation of tamoxifen injection, wounding experiments were conducted. Mice were anesthetized (90 μg ketamine plus 10 μg xylazine/g), shaved, depilated with Nair and cleaned with alcohol. Per mouse, four bilateral full-thickness skin wounds were created, using a sterile 4 mm biopsy punch on the dorsorostral back skin without injuring the underlying muscle. Wounds were separated by a minimum of 6 mm of uninjured skin, and digitally photographed at 0, 3, 7 and 10 days after wounding using a Sony D-9 digital camera. The wound area was determined using Northern Eclipse (Empix) software, and wound closure was expressed as percentage of initial wound size. In addition, mice were sacrificed using CO₂ at 3, 7 and 10 days after wounding for histological, immunohistochemical and hydroxyproline analyses.

Wound tissue sections (0.5 μm) were cut using a microtome (Leica, Richmond Hill, ON) and collected on Superfrost Plus slides (Fisher Scientific, Ottawa, ON). Sections were then de-waxed in xylene and rehydrated by successive immersion in descending concentrations of alcohol. The sections were then subjected for immunofluorescence staining. Briefly, tissue sections were blocked by incubated with 5% donkey serum for 1 hour and washed with phosphate-buffered saline (PBS). Sections were then incubated with primary antibodies for 1 hour at room temperature under humidified conditions. Primary antibodies used were: mouse anti-α-smooth muscle actin (anti-α-SMA; 1:500 dilution, Sigma), rabbit anti-proliferating cell nuclear antigen (anti-PCNA; 1:500 dilution, Abcam, Cambridge, MA). After primary antibody incubation, sections were washed with PBS and incubated with appropriate fluorescent secondary antibodies (Jackson ImmunoResearch, West Grove, PA) for 1 hour at room temperature. Sections were then washed with PBS, mounted with mounting medium containing DAPI and photographed using a Zeiss fluorescence microscope and Northern Eclipse software (Empix, Missassagua, ON). The tissue expression of α-SMA was graded on a scale of 0–3 by three blinded observers; 0 signifies no staining, 1 signifies very little staining, 2 signifies moderate staining, 3 signifies extensive staining. For PCNA staining, the percentage of positive cell staining was calculated per mm² using image analysis software (Northern Eclipse, Empix). To assess the effects of integrin β1 deletion on wound collagen synthesis, trichrome collagen stain was used. To assess the amount of granulation tissue, hematoxylin and eosin staining and Northern Eclipse software (Empix) was used.

To locally supply additional TGFβ1 in the wound vicinity, carrier-free recombinant human TGFβ1 (R&D Systems, Wiesbaden, Germany) was injected at four sites around the wound, allowing to infiltrate the wound margins, at a total dose of 0.40 μg as described (Peters et al., 2005). Mock injections were made using only the carrier PBS. First injections were made on day 1 after wounding, followed by further injections every second day until wounds were harvested for paraffin embedding at 7 days after wounding.

Hydroxyproline assay was performed essentially as described previously (Liu et al., 2008; Liu et al., 2009a; Reddy and Enwemeka, 1996). Wound tissues were homogenized in saline, hydrolyzed with 2N NaOH for 30 minutes at 120°C, followed by the determination of hydroxyproline by modification of the Neumann and Logan's reaction using Chloramine T and Ehrlich's reagent using a hydroxyproline standard curve and measuring at 550 nm. Values were expressed as μg hydroxyproline per mg protein.

Primary dermal fibroblasts were isolated from explants (4- to 6-week-old animals) as described (Chen et al., 2005). Cells were subjected to indirect immunofluorescence analysis as described (Chen et al., 2005; Kennedy et al., 2007) using anti-α-SMA, Rhodamine-phalloidin (Sigma) and anti-vinculin (Sigma) antibodies, followed by an appropriate secondary antibody (Jackson ImmunoResearch). Cells were photographed (Zeiss Axiphot B-100, Empix). Alternatively, cells were lysed in 2% SDS, proteins quantified (Pierce) and subjected to western blot analysis as described (Shi-wen et al., 2004; Chen et al., 2005; Kennedy et al., 2007). Anti-integrin-β1 antibody was from R&D Systems; anti-integrin-β3, anti-TGFβ type I receptor, anti-TGFβ type II receptor and anti-GAPDH antibodies were from Santa Cruz Biotechnology. Fluorescence intensity of α-SMA fibers was quantified by line scan measurement using Northern Eclipse (Empix) software (Liu et al., 2007). MLEC-PAI/Luc, a mink lung epithelial cell line stably transfected with an expression plasmid containing a TGFβ-responsive region (−799 to +71) of human plasminogen activator inhibitor-1 (PAI1) gene fused to the firefly luciferase reporter gene. Cells were maintained in DMEM supplemented with 5% fetal calf serum.

RT-PCR was performed essentially as described (Kennedy et al., 2007; Kennedy et al., 2008; Pala et al., 2008). A total of ten control and conditional-knockout mice were analyzed independently for each data point (mean ± s.d. shown from these ten independent animals; Student's paired t-test). Cells were cultured until 80% confluence, serum starved for 24 hours and total RNA was isolated (Qiagen). Integrity of the RNA was verified by gel electrophoresis. Total RNA (25 ng) was reverse transcribed and amplified (TaqMan Assays on Demand; Applied Biosystems) as described (One-step Mastermix; Applied Biosystems) using the ABI Prism 7900 HT sequence detector (Perkin-Elmer-Cetus, Vaudreuil, QC). Triplicates of each samples were run, and expression values were standardized to values obtained with control 18S RNA primers using the ΔΔC_t method.

Fibroblasts were isolated and cultured as described above. Adhesion assays were performed essentially as previously described (Chen et al., 2004; Chen et al., 2005). Wells of 96-well plates were incubated overnight at 4°C with 10 μg/ml fibronectin (Sigma) in 0.5% bovine serum albumin (BSA), 1 × PBS. Subsequently, cells were blocked for 1 hour in 10% BSA in PBS, room temperature. Fibroblasts were harvested with 2 mM EDTA in PBS (20 minutes, room temperature), washed twice with DMEM serum-free medium containing 1% BSA (Sigma), resuspended in the same medium at 2.5 ×10⁵ cells/ml and 100 μl suspension was incubated in each well for the times indicated. Non-adherent cells were removed by washing with PBS. To detect cell adhesion, an acid phosphatase assay was used; adherent cells were quantified by incubation with 100 μl substrate solution (0.1 M sodium acetate, pH 5.5; 10 mM p-nitrophenylphosphate and 0.1% Triton X-100) for 2 hours at 37°C. The reaction was stopped by the addition of 15 μl of 1 N NaOH per well and A₄₅₀ was measured. Comparison of adhesive abilities was performed by using Student's unpaired t-test. P<0.05 was considered statistically significant.

Experiments were performed essentially as described (Shi-wen et al., 2004). For a floating gel assay, 24-well tissue culture plates were pre-coated with BSA. Cells were used at passage three. Trypsinized fibroblasts were suspended in MCDB medium (Sigma) and mixed with collagen solution [one part 0.2 M HEPES, pH 8.0, four parts 3 mg/ml collagen (Nutragen, Inamed) and five parts of 2 × MCDB] for a final concentration of 80,000 cells per ml in 1.2 mg/ml collagen. Collagen cell suspension (1 ml) was then added to each well to polymerize. Gels were then detached from wells by adding 1 ml MCDB medium. Gel contraction was quantified by measuring changes in weight. When indicated, contraction assays were performed in the presence of DMSO and 100 μM blebbistatin (myosin ATPase/contraction inhibitor, Calbiochem).

Measurement of contractile forces generated within a three-dimensional FPCL was performed as described (Shi-wen et al., 2004). Briefly, using 1 ×10⁶ cells/ml of collagen gel (First Link, UK), we measured the force generated across a collagen lattice using a culture force monitor which measures forces exerted over 24 hours. A rectangular FPCL was cast, floated in medium in 2% FCS and attached to a ground point at one end and a force transducer at the other. Cell-generated tensional forces in the collagen gel are detected by the force transducer and logged into a personal computer (t-CFM) (Denton et al., 2009). Graphical readings are produced every 15 seconds providing a continuous output of force (Dynes: 1 ×10⁻⁵ N) generated. Experiments were run in parallel and three independent times, and a representative trace is shown.

For in vitro wounding (migration) experiments, fibroblasts obtained from integrin β1 conditional (C/C) and knockout (K/K) mice were cultured in 12-well plates (2 ×10⁵ cells/well). The next day, the cells were confluent. Medium was removed, and cells rinsed with serum-free medium with 0.1% BSA and cultured for an additional 24 hours in serum-free medium plus 0.1% BSA. The monolayer was artificially injured by scratching across the plate with a blue pipette tip (approximately 1.3 mm width). Cells were washed twice to remove detached cells or cell debris, and cultured in serum-free medium in the presence of mitomycin C (10 μg/ml, Sigma) to prevent cell proliferation. After 24 and 48 hours, images of the scratched areas under each condition were photographed. Migration distance was measured (mean ± s.d.) using Northern Eclipse software (Empix). For path trajectory studies, cells were plated at low density (5000 cells/cm²) on type I collagen (30 μg/ml in PBS, acid-extracted from newborn calf skin) or tissue culture plastic. Cells were observed for 5 hours in an incubator chamber attached to an IX81 microscope (Olympus Europa, Hamburg) at 37°C, 5% CO₂ and 60% humidity (n=69–123). Frames were taken every 30 minutes using an OBS CCD FV2T camera (Olympus). To quantify cell movements, trajectories were measured using OBS Cell R software (Olympus). Persistence was calculated as the ratio of net length to total length, where net length is distance between start and end point, and total length is length of the entire path (Danen et al., 2005). For the cell proliferation assay, cells were seeded in 96-well plates at 2000 cells/well. Cell number was determined after 24, 48 and 72 hours of incubation using the MTT kit (Roche).

Fibroblasts (80,000 cells per ml collagen) co-cultured by embedding within a 3D collagen gel matrix with mink lung cells to produce which luciferase under control of the PAI-1 promoter in response to TGF-β (Abe et al., 1994; Wipff et al., 2007). Activation of TGFβ was assessed 24 hours later by lysing mink lung cells, and luciferase activity was assessed by light production from a luciferin substrate (Promega) using a luminometer (Wallac). Data are presented as mean ± s.d. of one representative experiment performed in triplicate. Experiments were performed a minimum of three times. Co-cultures were performed in the absence and presence of a TGFβ neutralizing antibody [1D11; R&D Systems (10 μg/ml)], ALK-5 inhibitor [SB431542; Tocris (10 μM) a specific inhibitor for TGF-β type I receptor (TβRI)/ALK-5] or blebbistatin (100 μM; Calbiochem). As controls, cells were incubated with IgG (Jackson ImmunoResearch) or DMSO.

This work was supported by grants from the Canadian Foundation for Innovation, the Canadian Institutes of Health Research and the Ontario Thoracic Society (to A.L.), the Arthritis Research Campaign (UK) and the Scleroderma Society (to D.J.A. and C.P.D.), and the Deutsche Forschungsgemeinschaft (SFB829 to B.E. and T.K.). A.L. is a New Investigator of the Arthritis Society (Scleroderma Society of Ontario), the recipient of an Early Researcher Award and a member of the Canadian Scleroderma Research Group New Emerging Team.