TNAP limits TGF-β-dependent cardiac and skeletal muscle fibrosis by inactivating the SMAD2/3 transcription factors

Fibrosis accompanies many chronic diseases and is a common response to injury. It is characterized by excessive production, and accumulation of, collagen and other extracellular matrix (ECM) components, cellular dysfunction and loss of tissue architecture that eventually lead to organ failure (Ueha et al., 2012). Fibrosis can affect many organs, including heart and skeletal muscle, and has become a major cause of death in the developed world.

Fibrosis is an integral component of most cardiac pathological conditions and can be associated with cardiomyocyte death (Frangogiannis, 2012), pressure or volume overload (Berk et al., 2007), hypertrophic cardiomyopathy (Kania et al., 2009) and toxic insults (Bernaba et al., 2010).

Skeletal muscle fibrosis accompanies ageing, in which case gradual muscle loss leads to the accumulation of adipose tissue and ECM (Alnaqeeb et al., 1984; Wood et al., 2014). Fibrosis occurs early in congenital muscular dystrophies and impairs muscle function and regeneration (Serrano and Muñoz-Cánoves, 2017; Tedesco and Cossu, 2012). In both cardiac and skeletal muscle, fibrosis increases the stiffness of the tissue, thereby directly impairing muscle cell contraction.

The molecular pathways involved in fibrosis are well known and common for the two organs. Transforming growth factor β (TGF-β), in addition to being a major morphogen during development, is one of the main signalling molecules initiating fibrosis and is secreted by many cell types in the injured tissue. Its activity is controlled by proteolytic cleavage of a precursor, previously bound to inhibitory proteins and subsequently released by proteases. Active TGF-β binds to the serine threonine protein kinases receptors (termed TGF-β receptor type I and type II) and induces the phosphorylation and activation of SMAD2 and SMAD3 (SMAD2/3) transcription factors (also known as receptor-regulated SMADs or R-SMADs) (Heldin et al., 1997; Zi et al., 2012). The phosphorylation of R-SMADs triggers their association with SMAD4, leading to their nuclear translocation, where they control the transcription of pro-fibrotic genes. Finally, the dephosphorylation of R-SMADs promotes their exit from the nucleus and the end of the signalling (Bruce and Sapkota, 2012). As a central regulator of a number of physiologically important processes, TGF-β signalling is highly regulated. Cellular susceptibility is determined by receptor endocytosis via clathrin- and caveolin-dependent pathways (Di Guglielmo et al., 2003). Depletion of TGF-β ligand defines the transcriptional response and is achieved by TGF-β binding to the cell surface or its cellular uptake (Clarke et al., 2009). The duration and temporal dynamics of ligand availability influence downstream signalling response, as repeated pulses of TGF-β can lead to long-term sustained response (Zi et al., 2011). Several post-translational modifications are capable of regulating R-SMAD activity. These include ubiquitylation (Dupont et al., 2012, 2005; Episkopou et al., 2001; Lin et al., 2000; Zhu et al., 1999), acetylation (Grönroos et al., 2002), parpylation (Lönn et al., 2010), sumoylation (Miles et al., 2008) and phosphorylation at different serine residues (Kretzschmar et al., 1997). Rapid inactivation of R-SMADs can be achieved by dephosphorylation. However, to date only one R-SMAD phosphatase, protein phosphatase 1A (PPM1A), has been identified for the TGF-β pathway (Lin et al., 2006). Identification of new R-SMAD phosphatases would enable efficient attenuation of pro-fibrotic signalling in acute and chronic diseases.

Tissue non-specific alkaline phosphatase (TNAP, also known as ALPL) is an enzyme that is widely expressed in many organs. It is involved in bone mineralization and mutations in the TNAP-encoding gene have been associated with hypophosphatasia, a rare inherited disorder, characterized by defective bone mineralization (Mornet et al., 2001). Recently TNAP has been associated with vascular calcification (Romanelli et al., 2017; Sheen et al., 2015) and cardiac fibro-calcification (Herencia et al., 2015). Its role in fibrosis and in the context of TGF-β signalling has not been addressed.

In the present study, we demonstrate that TNAP negatively regulates TGF-β signalling pathway by regulating the level of phosphorylation of SMAD2/3. We show that TNAP is highly induced in hypertrophic hearts and dystrophic skeletal muscle, and that the abrogation of TGF-β signalling in TNAP-positive cells reduces vascular and interstitial fibrosis. TNAP interacts with SMAD2 and its overexpression decreases SMAD phosphorylation and the expression of downstream fibrotic genes.

We previously established TNAP expression in skeletal muscle and demonstrated that its high activity marks perivascular cells (Dellavalle et al., 2007). In contrast, we found that TNAP is more widely expressed in healthy mouse heart. We detected alkaline phosphatase activity mainly in the interstitium (Fig. 1A,B) but also in cardiomyocytes (Fig. 1C). At higher resolution (Fig. 1D) AP activity colocalized both with endothelial (VE-cadherin⁺) and perivascular (NG2⁺) cells and also extended outside of the small vessels (Fig. 1D). Of the three alkaline phosphatase isoforms, intestinal (ALPI), embryonic (ALPG) and non-specific (TNAP), only the latter is expressed in the heart at any developmental stage examined (Fig. 1E). When TNAP-Cre^ERT mice (Dellavalle et al., 2011) were crossed to the Rosa26NZG reporter mice, LacZ activity was detected in many areas of the heart (Fig. 1F). Additional examples of TNAP localization in adult healthy heart are shown in Fig. S1.

We next analysed changes in TNAP activity in response to fibrosis. We used a model of angiotensin II-induced cardiac hypertrophy, which is known to cause fibrosis after few days (Crowley et al., 2006). We implanted subcutaneous mini-pumps releasing angiotensin II (Ang II) (2.8 mg/kg of body weight per day) for 2 weeks in 10-week-old C57/BL6J mice. We analysed the hearts of treated mice for the presence of hypertrophy and fibrosis. As expected, the treatment with Ang II caused cardiomyocyte hypertrophy and led to an increase in tenascin C-positive areas compared with mice treated with saline solution (Sham) (Fig. S2a–i). Moreover, the expression of the fibrotic genes Fibronectin 1 (Fn1), tenascin C (Tnc), smooth muscle actin (Sma, also known as Acta2), vimentin (Vim) and collagen 1a2 (Col1a2) was significantly upregulated in Ang II-treated mice compared with the control (Fig. S2j–p).

We took advantage of this disease model to analyse the expression of Tnap. Interestingly, we found a significant increase in the number of TNAP-positive cells in hypertrophic hearts in comparison to the control (Fig. 2A–C; n=5 for each group). Moreover, we detected an upregulation of Tnap expression by quantitative (q)RT-PCR (Fig. 2D). In order to understand which cell types express TNAP in physiological and pathological conditions, we stained the heart sections for TNAP and laminin, vimentin and PECAM, which allow identification of cardiomyocytes basal lamina, cardiac interstitial and endothelial cells, respectively, and using confocal microscopy, counted the number of double-positive cells in a 77 mm² area. We confirmed that all three cells types analysed expressed TNAP in normal heart (TNAP/PECAM⁺ cells, 16.6±6%; TNAP/laminin⁺ cells, 26±5.4%; TNAP/vimentin⁺/PECAM⁻, cells 27±5.6%; n=3 mice, mean±s.d.). Remarkably, in the Ang II-treated mice the number of cells expressing TNAP significantly increased in all three different cell populations (TNAP⁺/PECAM⁺ cells, 30.5±4.1%; TNAP⁺/laminin⁺ cells, 42.4±1.5%; TNAP⁺/vimentin⁺/PECAM^– cells, 49.3±6%; n=3 mice, mean±s.d., P<0.001, two-tailed t-test) (Fig. 2E–M). These data prove that in hypertrophic hearts TNAP is significantly up regulated in cardiac myocytes, endothelial cells and interstitial cells.

However, it is important to notice that, under normal condition only a minority of cells in the heart express TNAP and even if this number almost doubles in the Ang II-treated animals, it remains confined to less than half of total cardiac cells.

Since the increased number of TNAP-positive cells is preferentially localized in tenascin C-positive regions (Fig. 2A,B,K,L), we reasoned that TNAP-positive cells might have a role in the regulation of fibrosis upon Ang II treatment. TNAP is widely expressed in many different cells, so that creating a cardiac-specific knockout would be technically very challenging in this context as Cre should be driven by all least three (cardiac, endothelial and interstitial) different promoters. Therefore, to prove that TNAP-expressing cells are in fact involved in cardiac fibrosis, despite being a minority of total cardiac cells, we inactivated TGF-β signalling specifically in TNAP-positive cells. To achieve this goal, we took advantage of an inducible conditional knockout Tnap^cre;Tgfbr2 floxed mouse (Chytil et al., 2002; Dellavalle et al., 2011). This strain is characterized by the inactivation, upon tamoxifen injection, of TGF-β receptor 2 (TGF-βR2) in TNAP-expressing cells. We first verified by immunofluorescence whether TNAP-positive cells express TGF-βR2 and found that virtually all TNAP-positive cells are also positive for TGF-βR2, both in Sham- and Ang II-treated mice (Fig. 3A,B). We then injected the Tnap^cre;Tgfbr2^fl/fl transgenic mice with 100 mg/kg of body weight of tamoxifen once a day for 5 days; at day 7 we implanted subcutaneous mini-pumps releasing 2.8 mg/kg of body weight per day of Ang II for 2 weeks (Fig. 3C). We verified Cre recombination efficiency on the total heart of the Tnap^cre;Tgfbr2^fl/fl mice by qRT-PCR for the exon 4 of the Tgfbr2, the exon excised by Cre recombinase (Chytil et al., 2002), and found an ∼50% reduction in its expression in Tnap^cre;Tgfbr2^fl/fl compared to control (Fig. 3D). We confirmed this recombination efficiency by counting the number of TNAP/tdTomato double-positive cells over the total number of TNAP-positive cells in the Tnap^cre;Rosa26-tdTomato mice (Fig. S3). Injection of control oil did not modify the expression of TNAP.

We next quantified the area of fibrosis in Tnap^cre;Tgfbr2^fl/fl Sham-treated mice, Tgfbr2^fl/fl mice treated with Ang II and Tnap^cre;Tgfbr2^fl/fl mice treated with Ang II. We found a strong reduction of interstitial and vascular fibrosis in the Tnap^cre;Tgfbr2^fl/fl Ang II treated mice in comparison with the Tgfbr2^fl/fl mice treated with Ang II (Fig. 3E–L). This result was confirmed by Azan–Mallory staining and by immunofluorescence analysis, showing reduced expression of collagen 1a2 and MMP1 in Tnap^cre;Tgfbr2^fl/fl Ang II-treated mice in comparison with the Tgfbr2^fl/fl treated with Ang II (Fig. S4a-i). We analysed cardiac hypertrophy in Tnap^cre;Tgfbr2^fl/fll mice treated with Ang II compared with the control by calculating the heart weight:body weight ratio and by measuring the area of cardiomyocytes, but we found no difference in cell size and in overall hypertrophy between the two groups (Fig. 3M–Q), indicating that the reduction of fibrosis is not accompanied by a reduction in hypertrophy. We next characterized the expression of fibrotic genes by qRT-PCR in the three sample groups and found a significant reduction in collagen 1a2 and vimentin expression in the Tnap^cre;Tgfbr2^fl/fl mice treated with Ang II compared with the controls (Fig. 3R,S), while Sma expression did not change (Fig. 3T). Moreover, also immunostaining with antibodies against collagen 1 and MMP1 showed a dramatic reduction of areas of fibrosis, as also confirmed by Azan–Mallory staining (Fig. S4). These data prove that TNAP is specifically expressed and upregulated in cells with active TGF-β signalling that underlie cardiac fibrosis. It is remarkable that a 50% reduction of expression in a minority of the cardiac cell population leads to a dramatic decrease in the level of fibrosis, suggesting a specific role of TNAP-positive cells in regulating fibrosis.

Since it has been previously shown that TNAP is upregulated in dystrophic muscles (Diaz-Manera et al., 2012), we asked whether the abrogation of TGF-β signalling in TNAP-positive cells in dystrophic mice could reduce skeletal muscle fibrosis. We took advantage of the α-sarcoglycan deficient (Sgca-null) mice, a model for limb-girdle muscular dystrophy 2D (Duclos et al., 1998), which develop severe fibrosis after 2 months of age. We generated a triple transgenic strain by crossing the Sgca-null mice with the Tnap^cre;Tgfbr2^fl/fl mouse. We injected 6–7-day-old pups with tamoxifen for 3 days and killed the mice when they were 2 months old (Fig. 4A). We measured the recombination efficiency of Cre recombinase by qRT-PCR and western blotting (Fig. 4B,C) and found it to be 50%. We next analysed the extent of fibrosis by measuring the percentage of tenascin C-positive regions and found a significant decrease in fibrosis in the tibialis anterior (TA), the gastrocnemius (GA) and the diaphragm (DIA) of the Sgca-null Tnap^cre;Tgfbr2^fl/fl mice compared with the Sgca-null controls (Fig. 4D–G). We analysed the level of collagen 1a2 by immunofluorescence (Fig. 4H) and the expression level of collagen 1a2, Sma and vimentin mRNA by qRT-PCR (Fig. 4I–K). We observed a significant reduction of their expression in the Sgca-null Tnap^cre;Tgfbr2^fl/fl mice compared with what was found in the Sgca-null controls. In order to reveal whether the reduction of fibrosis in the triple transgenic mice is sufficient to improve muscle strength and function, we measured muscle grip strength by placing mice on an upside down grid (Bonetto et al., 2015). Wild-type mice started to explore the environment and spent little time in the same location. In contrast Sgca-knockout mice spent 26% of the total time without moving, which importantly decreased to only 10% in the triple transgenic mice (Fig. 4L) (n=4 for each group). These data suggest that the reduction in fibrosis is associated with improved skeletal muscle function. Overall, the data so far demonstrate that cells that express TNAP in fibrotic cardiac and skeletal muscle are of pivotal significance in the TGF-β-dependent development of the pathology as inhibition of TGF-β signalling in these cells is sufficient to limit fibrosis.

Since we obtained similar results in cardiac and skeletal muscle models of fibrosis, we hypothesized that TNAP may have a role in the regulation of TGF-β signalling and that this role is conserved in both tissues. Being a phosphatase, TNAP could be involved in the dephosphorylation of phosphorylated (p-)SMAD2/3 and have thereby a regulatory role in the transcription of fibrotic genes. In order to test our hypothesis, we first verified whether TNAP and p-SMAD2/3 colocalize within the cell. We stained sections for TNAP and p-SMAD2/3 and saw that both of them partially colocalize in the peri-nuclear region (Fig. 5A). We confirmed their colocalization also in vitro. We transfected C2C12 skeletal myogenic cells with EYFP-SMAD2 and mCherry-TNAP plasmids and confirmed that the two proteins colocalize in intracellular vesicular compartments around the nucleus (Fig. 5B). Interestingly TNAP and SMAD2 colocalization was present in both Sham and Ang II treated mice (Fig. 5A). We found that TNAP not only colocalizes with SMAD2 but also physically interacts with it. We proved this by performing a proximity ligation assay (PLA) (Shah et al., 2017) (Fig. 5C) and co-immunoprecipitation for C2C12 cells transfected with Myc–TNAP (Fig. S5).

Our results show that endogenous SMAD2 interacts with Myc–TNAP. Notably, the TNAP–SMAD2 interaction not only occurs upon TGF-β stimulation, but also, at a lower level, under control conditions. In order to understand whether this interaction has a functional relevance on TGF-β signalling, we treated cardiac fibroblasts with TGF-β1 (5 ng/ml) and the specific TNAP inhibitor MLS-0038949 (Crowley et al., 2006) (200 nM) for 1 h and analysed SMAD2/3 phosphorylation by western blotting. The cells treated with TGF-β1 and TNAP inhibitor showed a significant increase in SMAD2/3 phosphorylation compared with cells treated with only TGF-β1 (Fig. 5D) (n=3 independent experiments), suggesting that the blockade of TNAP leads to enhanced TGF-β signalling. To provide further evidence for TNAP regulation of SMAD2/3 activity, we overexpressed plasmid encoding for Myc–TNAP in C2C12 skeletal muscle cells and stimulated the cells with TGF-β1 (5 ng/ml). We chose this cell line because it has no detectable TNAP expression (Fig. 5E). We verified that the plasmid was functional by performing a functional staining assay for AP in the C2C12 cells (Fig. 5F,G). Upon TGF-β1 stimulation, we detected a decrease in SMAD2/3 phosphorylation in Myc–TNAP overexpressing cells compared with the GFP control (Fig. 5H,I). Inhibition of SMAD2/3 phosphorylation was not dramatic because on average ∼20–30% of C2C12 cells are transiently transfected. Together, these data prove that TNAP is capable and in itself sufficient to limit SMAD2/3 phosphorylation. We next analysed the expression of fibrotic genes downstream p-SMAD2/3 by RT-qPCR and observed that expression of fibronectin 1, collagen 1a2 and tenascin C mRNAs were downregulated in Myc–TNAP-overexpressing cells compared with the control (Fig. 5J), indicating that TNAP-induced dephosphorylation of p-SMAD2/3 is sufficient to limit transcription of fibrotic genes. Taken together, these data indicate that TNAP regulates SMAD2/3 phosphorylation levels and is a direct attenuator of cellular pro-fibrotic signalling pathways.

Despite ever increasing understanding of the molecular processes that trigger fibrosis across different organs, there are only a few therapies available and most of them only indirectly target fibrotic processes (Rockey et al., 2015). TGF-β1 binding to the receptors on the plasma membrane leads to the phosphorylation of R-SMADs, their translocation into the nucleus and transcription of pro-fibrotic genes. Given the importance of TGF-β signalling and the phosphorylation of R-SMADs in fibrosis, the identification of phosphatases, able to switch off the TGF-β pathway, has been of interest for the past ten years since they could also be targets for anti-fibrotic drugs. Nevertheless, until now, only one phosphatase, PPM1A, has been identified (Lin et al., 2006). Here, we show that TNAP is able to dephosphorylate SMAD2/3 and consequently limit the transcription of pro-fibrotic genes.

An abundant literature exists on TNAP function in bone development and calcification (Byon et al., 2008; Millan, 2013); however, the role of TNAP in the heart has only been partly elucidated. TNAP overexpression has recently been associated with increased cardiac fibrosis and vascular calcification. Sheen et al. demonstrated that the induced expression of TNAP in vascular smooth muscle cells is sufficient to cause medial vascular calcification (Sheen et al., 2015). Similarly, Romanelli and colleagues showed that TNAP overexpression in vascular endothelium in mice leads to arterial calcification and subsequent coronary atherosclerosis (Romanelli et al., 2017). These studies described the effects of sustained TNAP overexpression, but the role of TNAP under physiological levels of expression has not been addressed. Indeed, it is counterintuitive that tissues where calcification is exclusively pathological have maintained TNAP expression during evolution. This suggests that TNAP may also play beneficial roles in striated muscle that is eventually overridden by sustained fibrotic stimuli; however, this role has thus far remained unknown. Here, we analysed two different models of fibrosis: rapid-onset cardiac fibrosis induced by the hypertensive molecule Ang II, and chronic skeletal muscle fibrosis in the Sgca-null mice. We confirm earlier evidence showing that TNAP is upregulated in tissue fibrosis (Herencia et al., 2015; Gan et al., 2014), but also show here that the number of cells expressing TNAP increases in both heart and skeletal muscle. Furthermore, as the abrogation of TGF-β signalling in TNAP-positive cells dramatically reduces vascular and interstitial fibrosis, we can conclude that TNAP is actively expressed in cells driving tissue fibrosis. Since in the heart, not only interstitial and perivascular cells, but also endothelial cells and cardiomyocytes participate in the process (though likely to a minor extent), it is interesting to note that TNAP has a similar pattern of expression, being more abundant in interstitial and perivascular cells but also present in the other cell types of the heart.

TGF-β signalling relies on trafficking of activated phosphorylated R-SMAD proteins from the plasma membrane to the nucleus (Chen, 2009; Hough et al., 2012; Nakao et al., 1997). SMAD2/3 inactivation during this process would enable a cell to rapidly limit the expression of pro-fibrotic genes even in the presence of TGF-β secreted in the inflammatory tissue environment. Here, we show that TNAP inhibition enhances SMAD2/3 phosphorylation whereas TNAP overexpression has the opposite effect. We demonstrate that TNAP interacts with SMAD2. Our data collectively prove that TNAP is an efficient inhibitor of SMAD2/3 and limits SMAD-dependent transcription of pro-fibrotic genes. Intriguingly, TNAP interacts with SMAD2 even in cells that are not exposed to TGF-β1. These results demonstrate that TNAP likely has a housekeeping function during normal tissue growth and homeostasis, which explains its basal level expression in cardiac and skeletal muscle. In addition to the molecular mechanism described here, TNAP may also indirectly affect the activity of numerous phosphorylation-dependent pathways as this enzyme is known to regulate the level of cytoplasmic inorganic phosphate (Pi) (Hessle et al., 2002).

In conclusion, we show here that the ubiquitously expressed TNAP is an essential regulator of SMAD2/3-dependent TGF-β signalling in physiological conditions and in tissue fibrosis. These results provide a novel insight into the mechanisms that control the expression of extracellular matrix genes and may lead to the development of novel therapies for fibrosis.

Mice were maintained in pathogen-free conditions at Manchester University animal facility and fed standard rodent diet. Animal studies were performed according to the United Kingdom Animals (Scientific Procedures) Act 1986 and approved by the University of Manchester Ethics Committee. Angiotensin experiments were approved under Home Office license no. 40/3625 and no. P3A97F3D1. All the other experiments were approved under licence no.707435. Sgca-null mice (Duclos et al., 1998), Tnap^cre mice (Dellavalle et al., 2011) and Tgfbr2^flox/flox (Chytil et al., 2002) mice were backcrossed onto C57BL/6 mice (Charles River Laboratories Inc., Wilmington, MA) and were genotyped as previously described (Dellavalle et al., 2011). Rosa26NZG mice were obtained from the Jackson laboratory. Mice were culled by administration of isoflurane and consequent cervical dislocation at appropriate time points. Hearts and muscles were collected in ice-cold PBS and then embedded in glue or OCT inclusion medium. Samples were stored at −80°C before processing.

In order to induce cardiac hypertrophy, angiotensin II (Sigma-Aldrich, St. Louis, MO, USA, Cat. No. A9525) (2.8 mg/kg of body weight/day) or vehicle (ddH₂O) was administrated to 10-week-old wild type (wt) and transgenic male mice via osmotic mini-pumps (1002, Alzet, Cupertino, CA) implanted subcutaneously. Briefly, the mice were anaesthetized with isoflurane (5% induction; 2–3% maintenance) in oxygen, using a small nose cone attached to a modified Bane circuit. The skin on the dorsal surface of the animal was shaved with a clipper, and sterile scissors were used to make a 0.5 cm horizontal incision. Subcutaneous tissue was spread using a haemostat to create a pocket in which the minipump was inserted. The incision was closed with absorbable suture (5-0 Maxon). After the surgery the mice received 0.1 mg/kg of body weight of buprenorphine and were allowed to recover before being returned to the home cage. Hearts were isolated and analysed for cardiac hypertrophy following 14 days of Ang II infusion. In order to induce the expression of the Cre recombinase in the null Tnap^cre Tgfbr2^fl/fl and Sgca^−/− Tnap^cre;Tgfbr2^fl/fl mice, tamoxifen (Sigma-Aldrich, Cat. No. T5648) was dissolved in corn oil (Sigma-Aldrich, Cat. No. C8267) and injected at the following concentrations and time points: 0.3 mg/mouse in 6-day-old pups once a day for 3 consecutive days, and 3 mg/mouse in 10-week-old mice once a day for 5 consecutive days.

The four limb-hanging test (also known as Kondziella's inverted screen test) (Bonetto et al., 2015) represents a method to assess muscle strength using all four limbs and to determine the general condition over time. In this study, we modified this method to improve its reproducibility. Briefly, we placed the mice upside down on a grid (50 cm×30 cm) and we counted the number of seconds in which the mice were not moving on the grid, over a total period of 50 s each mice. We calculated the resting time on the grid as the average of three different tests for each mouse, and we analysed four mice per each group.

PLA was performed according to the manufacturer's instructions using the Duolink In Situ Orange Kit Goat/Rabbit (Sigma-Aldrich, Cat. No. DUO92106). Immunofluorescence was performed as previously described (Arn et al., 2014). Briefly sections (5 µm) were fixed in 4% paraformaldehyde (Sigma-Aldrich, Cat. No. P6148) in PBS, pH 7.2 for 20 min; then washed three times, 5 min each, in PBS and the blockage of non-specific binding was undertaken with blocking buffer [1× PBS, 10% fetal bovine serum (FBS), 1 mg/ml bovine serum albumin (BSA), 0.1% Triton X-100], for 1 h at room temperature. For all steps, slides were kept in humidified chambers. For the TNAP staining, the sections were additionally fixed with methanol-acetone for 10 min. Antibodies were diluted in blocking buffer and incubated at +4°C overnight as suggested by manufacturer's instructions. The following day, sections were rinsed in PBS three times for 5 min each time, and fluorescent secondary antibodies, (Alexa Fluor conjugated, Thermo Fisher Scientific) diluted in blocking buffer were applied according to the manufacturer's instructions. Slides were washed three times in PBS for 5 min each time and incubated in DAPI (Sigma-Aldrich, Cat. No. 10236276001) for nuclei counterstaining. The following antibodies were used: goat anti-TNAP (Santa Cruz Biotechnology, Dallas, TX, Cat. No. c-23430), 1:100; rabbit anti-TNAP (Abgent, Suzhou city, China, Cat. No. AP1474C), 1:100; rat anti-PECAM (Dev. Studies Hybridoma Bank), 1:3; rabbit anti-laminin (Santa Cruz Biotechnology, Cat. No. sc-17810), 1:100; chicken anti-GFP (Abcam, Cambridge, UK, Cat. No. ab13970), 1:500; rabbit anti-collagen 1a2 (Merck Millipore, Billerica, MA, Cat. No. AB 745), 1:75; rabbit anti-tenascin C (Merck Millipore, Cat. No. AB19013), 1:150; chicken anti-vimentin (Novus biologicals, Littleton, CO, Cat. No. NB300-223) 1:300; rabbit anti -TGF-βR2 (Novus Biologicals, Cat. No. NB100-91994), 1:100; rabbit anti-SMAD2 (Cell Signalling, Danvers, MA, Cat. No. D43B4), 1:100; rabbit anti-pSMAD2/3 (Cell Signalling, Cat. No. D27F4), 1:100; and rabbit anti-Ki67 (Abcam, Cat. No. ab15580), 1:200. Appropriate fluorophore-conjugated secondary antibodies (Alexa Fluor 488, 594 and 647, Molecular Probes) were applied. Nuclei were visualized using DAPI. For the detection of TNAP activity, the Red Alkaline Phosphatase (Red AP) substrate kit (VECTOR Laboratories, Peterborough, UK, SK-5100) was used according to manufacturer's instructions. Light (Zeiss Axio Imager M2, with 5×, 10× and 20× objectives) and confocal (Leica, SP5 with ×40 and ×63 objectives) microscopy was performed to analyse tissue and cell staining. For confocal imaging, pictures were acquired at 1× line average and 3× frame average on a single confocal plane. Analysis was performed by using Zeiss ZEN 2 software and Adobe Photoshop CS5 Version 12.0×64 software. Quantification of tenascin C-positive regions in hearts and skeletal muscles was performed by using ImageJ 2.0.0-rc-9.

C2C12 cells from European Collection of Authenticated Cell Cultures, Public Health England (Porton Down, UK) were cultured in DMEM (Gibco, Thermo Fisher Scientific, Waltham, MA), supplemented with 1% L-glutamine, 10% fetal bovine serum (FBS), 1% penicillin-streptomycin (Thermo Fisher Scientific, Cat. No. 15140122). Cells were transfected with Lipofectamine 2000 (Thermo Fisher Scientific, Cat. No. 11668027) according to manufacturer's instructions. The following plasmids were used for transfection: pEGFP-N1 (Clontech, Cat. No. 6085-1), pCDNA3.1 N-Myc-TNAP and pCDNA3-mCherry TNAP. In detail, the Smad2/TNAP ORF was RT-PCR amplified from mouse cDNA. PCR products were inserted into EcoRI/XhoI site of the EYFP/mCherry/myc-pcDNA3 vectors.

Cardiac fibroblasts were obtained from neonatal and adult C57Bl/6 mice. Mice were killed by cervical dislocation and their hearts were rapidly removed. The hearts were then mashed with scalpels and digested with 10 ml collagenase solution [120 mg collagenase A (Roche, Cat. No. 10 103 578 001) and 12 mg protease (Sigma-Aldrich, Cat. No. 10165921001) dissolved in 80 ml PBS solution] at 37°C for 15 min, for five times. Cells were collected after each digestion and the collagenase was deactivated by addition of 50% FBS. The fibroblasts were centrifuged for 5 min at 220 g. The cell pellet was resuspended in 10ml ACF medium (80% DMEM, 20% FBS, 1% penicillin/streptomycin and 1% non-essential amino acids). Fibroblasts were then plated in 10 ml delta-NUNC tissue culture plates (Thermo Fisher Scientific, Cat. No. 150318) overnight. The next day, the medium was removed and replaced with 10 ml ACF medium. C2C12 and cardiac fibroblasts were serum starved for 24 h before adding medium containing 5 ng/ml recombinant TGF-β1 for 2 days (R&D Systems, Minneapolis, MN, Cat. No. 7666-MB-005).

Hearts and skeletal muscles from wild-type, Tnap^cre;Tgfbr2^fl/fl and α-sarcoglycan-null Tnap^cre;Tgfbr2^fl/fl and controls were homogenized with the homogenizer SHM1 (Stuart Equipment Stone, Staffordshire, UK, Cat. No. 1171631) in RIPA lysis buffer [50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, supplemented with phosphatase and protease inhibitor cocktail (Sigma-Aldrich, Cat. No. PCC 1010)]. Cold lysis buffer (RIPA) was added into the wells after removing the media and washing the cells with PBS. The cells were scraped into sterile, pre-cooled tubes. The samples were incubated (with rotation) for 30–60 min at 4°C. Soluble proteins were quantified with a Bradford Protein Assay Kit (Bio-Rad, Cat. No. 5000006). 50 µg of protein extract was subjected to SDS-PAGE and subsequently blotted on to a nitrocellulose membrane. For immunoprecipitation experiments, C2C12 cells were collected in lysis buffer (50 mM Tris-HCl pH 7.8, 150 mM NaCl, 1 mM CaCl₂ and 1% Triton X-100) supplemented with protease inhibitor cocktail (Sigma-Aldrich, Cat. No. P8340). 1 mg/ml of pre-cleared lysates were incubated with anti-Smad2 (Cell Signalling, Cat. No. D43B4) at a dilution of 1:50 or the negative control rabbit IgG in lysis buffer with rotation overnight at +4°C. The samples were then incubated with protein A magnetic beads (Merck Millipore, Cat. No. LSKMAGA10) for 1 h at +4°C (with rotation) and the immunocomplexes were detected by western blotting. The following primary antibodies were used: goat anti-Myc (Abcam, Cat. No. ab9132), 1:1000; rabbit anti-TGF-βRII (Novus Biologicals, Cat. No. NB100- 91994), 1:500; mouse anti-vinculin (Santa Cruz Biotechnology, Cat. No. H300), 1:6000; rabbit anti-SMAD2 (Cell Signalling, Cat. No. D43B4), 1:1000; rabbit anti-pSMAD2/3 (Cell Signalling, Cat. No. D27F4), 1:1000. Secondary antibodies (Dako Donkey anti-goat P0449; Dako Goat anti-rabbit P0448) were incubated for 1 h at room temperarure, and signals were revealed using Millipore ECL kit (Cat. No. WBKLS0500). For quantitative measurement, blots were analysed with ImageJ 2.0.0-rc-9, normalizing band intensities to vinculin levels.

Total RNA from heart, skeletal muscle and cells was extracted by homogenizing the samples in Trizol (Thermo Fisher Scientific, Cat. No. 15596018) reagent following the standard manufacturer's procedure. Chloroform was added to the samples, which were incubated for 15 min at room temperature and then centrifuged at 12,000 g for 15 min. The aqueous layer was mixed with isopropanol, incubated for 10 min and centrifuged for 10 min at 12,000 g. 75% of ethanol was added and the samples were centrifuged at 12,000 g for 15 min. After the elimination of ethanol, the RNA was dissolved in 50 µl of water. DNase (Thermo Fisher Scientific, Cat. No. 18068015) digestion was done on all the samples. RNA was quantified with a Nanodrop 2000 and equalized to give the same concentration before cDNA synthesis. cDNA synthesis was performed using ThermoScript RT-PCR System (Thermo Fisher Scientific, Cat. No. K1622) according to the manufacturer's instructions in final volume of 20 µl. qRT-PCR was carried out using FastStart Essential DNA Master Mix (Cat. No. 06402712001; Roche diagnostics GmbH, Mannheim, Germany). qRT-PCR was performed in five biological replicates, using a standard dilution series on every plate. Samples were normalized to the values for the housekeeping genes Gapdh and Rlp19 with the following primers: Gapdh F, 5′-AGGTCGGTGTGAACGGATT-3′, Gapdh R, 5′-TGTAGACCATGTAGTTGAG-3′ and Rpl19 F, 5′-ATGAGTATGCTCAGGCTACAGA-3′ and Rpl19 R, 5′-GCATTGGCGATTTCATTGGTC-3′. Specific primers were used for gene expression analysis were: tenascin C, F, 5′-ACGGCTACCACAGAAGCTG-3′; R, 5′-ATGGCTGTTGTTGCTATGGCA-3′; Fibronectin 1, F, 5′-TTCAAGTGTGATCCCCATGAAG-3′; R, 5′-CAGGTCTACGGCAGTTGTCA-3′; vimentin, F, 5′-CGGCTGCGAGAGAAATTGC-3′; R, 5′-CCACTTTCCGTTCAAGGTCAAG-3′; Tgfbr2 exon4/5, F, 5′-GACCTCAAGAGCTCTAACATCC-3′; R, 5′-CTAGAACTTCCGGGGCCATG-3′; Tgfbr2, F, 5′-CCGCTGCATATCGTCCTGTG-3′; R, 5′-AGTGGATGGATGGTCCTATTACA-3′; Sma, F, 5′-CCCAGACATCAGGGAGTAATGG-3′; R, 5′-TCTATCGGATACTTCAGCGTCA-3′; Col1a2, F, 5′-GGAGAGAGGAGTCGTTGGAC-3′; R, 5′-GTTCACCCTTCACACCCTGT-3′; Gapdh, F, 5′-TTCACCATGGAGAAGGC-3′; R, 5′-GGCATGGACTGTGGTCATGA-3′; Tnap, F, 5′-GTGGATACACCCCCCGGGGC-3′; R, 5′-GGTCAACCTTCCCCCCAATGCA-3′; Alpi, F, 5′-ATGATGCCAACCGAAACCC C; R, 5′-GCGTGTCCTTCTCATTGGTAA-3′; Alpg, F, 5′-GAGCGTCATCCCAGTGGAG-3′; R, 5′-GGTAGCGGTTACTGTAGACACC-3′.

Data are expressed as the mean±s.d. or mean±s.e.m. of independent experiments. Comparisons were made using an unpaired t-test. Statistical tests were carried out using PRISM6.0e (GraphPad Software, La Jolla, CA). P<0.05 was considered statistically significant.