AMP-activated Protein Kinase Inhibits Transforming Growth Factor-β-induced Smad3-dependent Transcription and Myofibroblast Transdifferentiation*

In wound healing, myofibroblast transdifferentiation (MFT) is a metaplastic change in phenotype producing profibrotic effector cells that secrete and remodel the extracellular matrix. Unlike pathways that induce MFT, the molecular mechanisms that negatively regulate MFT are poorly understood. Here, we report that AMP-activated protein kinase (AMPK) blocks MFT in response to transforming growth factor-β (TGFβ). Pharmacological activation of AMPK inhibited TGFβ-induced secretion of extracellular matrix proteins collagen types I and IV and fibronectin. AMPK activation also prevented induction of the myofibroblast phenotype markers α-smooth muscle actin and the ED-A fibronectin splice variant. AMPK activators did not prevent MFT in cells transduced with an adenovirus expressing dominant negative, kinase-dead AMPKα2. Moreover, AMPK activators did not inhibit MFT induction in AMPKα1,2–/– fibroblasts, demonstrating a requirement for AMPKα expression. Adenoviral transduction of constitutively active AMPKα2 was sufficient to prevent TGFβ-induced collagen I, α-smooth muscle actin, and ED-A fibronectin. AMPK did not reduce TGFβ-stimulated Smad3 COOH-terminal phosphorylation and nuclear translocation, which are necessary for MFT. However, AMPK activation inhibited TGFβ-induced transcription driven by Smad3-binding cis-elements. Consistent with a role for AMPK in transcriptional regulation, nuclear translocation of AMPKα2 correlated with the appearance of active AMPKα in the nucleus. Collectively, these results demonstrate that AMPK inhibits TGFβ-induced transcription downstream of Smad3 COOH-terminal phosphorylation and nuclear translocation. Furthermore, activation of AMPK is sufficient to negatively regulate MFT in vitro.

Myofibroblast transdifferentiation (MFT) 2 is a fundamental cellular program activated in embryonic development, tumorstroma interactions, wound repair, and fibrosis in the lung, heart, liver, and kidney (1)(2)(3)(4). Myofibroblasts are mesenchymal cells that display phenotypic markers of both muscle and nonmuscle cells. At sites of tissue injury, MFT produces profibrotic effector cells that assemble a fibrotic lesion (1,3,4). MFT is induced in an autocrine or paracrine fashion by cytokines, growth factors, and mechanical forces such as shear stress (3,5).
The temporal sequence of MFT in wound healing is not entirely clear, but it involves early de novo expression of the ED-A fibronectin (ED-A FN) splice variant and formation of a proto-myofibroblast (1,6). The transition of proto-myofibroblasts into myofibroblasts is marked by abundant de novo expression of ␣-smooth muscle actin (␣SMA) (7). Stress fibers rich in ␣SMA develop and link to focal adhesions at the plasma membrane (1,2). Finally, the mature myofibroblasts secrete a fibrillar collagen-and fibronectin-rich extracellular matrix that is remodeled into a fibrotic lesion (3,8).
AMP-activated protein kinase (AMPK) was initially characterized as a protein activated by nutrient and bioenergetic stress that raises intracellular AMP and lowers ATP (9 -11). AMPK belongs to the evolutionarily conserved SNF1 (sucrose non-fermentor) family of serine/threonine kinases that regulate energy homeostasis in eukaryotes (12,13). The kinase exists as a heterotrimer with a catalytic ␣-subunit and regulatory ␤and ␥-subunits. AMPK is highly sensitive to small increments in AMP and to elevations in the AMP/ATP ratio (12,14). Stimulation of AMPK also requires phosphorylation of a critical threonine residue (Thr-172) in the activation loop of the ␣-subunit. At least two upstream AMPK kinases have been identified, including the tumor suppressor LKB1 and the Ca 2ϩ -calmodulin-dependent protein kinase kinases ␣ and ␤ (11,15,16). In cells, AMPK activation slows metabolic reactions that consume ATP and stimulates reactions that produce ATP, thereby restoring the AMP/ATP ratio and the normal cellular energy state (12). Recently it has become clear that AMPK can be activated independent of changes in the AMP/ATP ratio (15)(16)(17), implying that AMPK might have cellular functions unrelated to its role as a nutrient stress defense pathway.
In this study we asked whether MFT is negatively regulated by AMPK. Numerous stimuli of MFT have been identified, such as endothelin-1 and transforming growth factor-␤ (TGF␤), but negative regulation of MFT has not been extensively characterized. Here, we report that AMPK activation blocks Smad3-dependent transcription and MFT in response to TGF␤.

EXPERIMENTAL PROCEDURES
Experimental Reagents-We purified and characterized recombinant human adiponectin from the supernatant of 293-T cells (generous gift from Philipp Scherer, Albert Einstein College of Medicine) stably transfected with the pFM1 vector containing an Acrp30 gene exactly as described (18). Adiponectin from a mammalian tissue culture system was used because bacteria cannot properly process the hydroxylprolineated collagenous domain of adiponectin (18). Recombinant human TGF␤1 was from R&D, and AICAR was from Sigma. A-769662 has been previously described (19). Antibodies were as follows: affinity-purified rabbit anti-human collagen type I (Biodesign International); collagen IV and fibronectin (Rockland); ␣SMA (Epitomics); mouse monoclonal ED-A FN and ␤-actin (Sigma). All other antibodies were from Cell Signaling.
Cell Culture and Induction of MFT-Human primary mesangial cells (HMC) were from Cambrex Bioscience Inc. (Walkersville, MD) and were characterized and cultured as previously reported (20,21). Human primary mesangial cells were used in passages 4 -10. Where indicated we also studied MFT in murine embryonic fibroblasts null for AMPK ␣1,␣2 (22). To induce MFT, cells in 60 mm plates at ϳ80% confluence were made quiescent for 24 h in DMEM with 0.5% FBS, then stimulated with 0.5 ng/ml TGF␤-1. The media contained 50 g/ml ␤-aminopropionitrile to minimize crosslinking. Twelve, 24 or 48 h after adding TGF␤, the supernatant was collected and the monolayer was solubilized in a 5 M guanidine-0.1 M Tris buffer (pH 8.6) with protease inhibitors as described (23).
Adenoviral Transduction of Human Primary Mesangial Cells with Dominant Negative and Constitutively Active AMPK␣2-Replication-defective adenoviral vectors encoding dominant negative AMPK␣2 (Ad.dnAMPK␣2), constitutively active AMPK␣2 (Ad.caAMPK␣2) and control adenovirus (Ad.trGFP) have been described (24 -26). Adenovirus expressing a dominant negative mutant of AMPK ␣2 was provided by Dr. Ken Walsh (25,27). All viral constructs were amplified in 293 cells and purified by using an Adeno-X TM virus mini-purification kit from Clontech Inc. (Mountain View, CA). Viral titers were determined as plaque forming units. For transduction, human primary mesangial cells were incubated with adenovirus at a multiplicity of infection of 50 for 12 h. The optimal multiplicity of adenoviral infection was determined using a GFP-encoding adenovirus. Under these conditions, the transduction efficiency was Ͼ90%.
Direct ELISA for Collagens, Fibronectins, and ␣SMA-ELISA for secreted collagen I, collagen IV, and FN in the medium was as described (24,28). We also used ELISA to measure cell-associated ED-A FN and ␣SMA (29). 100 l of media or cell extracts were absorbed to Nunc Maxisorp 96-well plates overnight at 4°C. After washing three times with phosphatebuffered saline/0.1%Tween, nonspecific binding was blocked in 1.0% bovine serum albumin, 100 mM phosphate buffer, pH 8.2. The wells were then incubated with a primary antibody, the appropriate affinity-purified and biotin-conjugated goat anti-IgG, and a horseradish peroxidase-conjugated streptavidin. A solution of 3,5,2Ј,5Ј-tetramethylbenzidine was added followed by quenching with acid and an absorbance reading at 450 nm in a SpectraMax 190 (Molecular Devices). All values were in the linear range and were normalized for cell number.
Immunocytochemical and Immunofluorescent Analysis of ED-A FN and PSMAD3-For immunocytochemistry, cells were fixed in 2% paraformaldehyde for 20 min, incubated with 0.6% hydrogen peroxide for 30 min at room temperature, and permeabilized with 0.1% Triton X-100 in TBS for 15 min. After blocking with 3.0% bovine serum albumin in TBS-T for 1 h, primary antibody (1:500) was added overnight at 4°C. The primary antibody was detected using the ImmPress Universal Antibody detection system (Vector Laboratories) exactly as described by the manufacturer. The monolayer was mounted in Vectamount (Vector Laboratories). A similar protocol without hydrogen peroxide pretreatment was used for immunofluorescent staining with detection by a fluorescein isothiocyanateconjugated secondary antibody, counterstaining with 4Ј,6-diamidino-2-phenylindole and mounting in Vectashield. The PSmad3 antibody recognizes Smad3 phosphorylated at Ser-423/425. Images were acquired with a SPOT RT camera (Diagnostic Instruments). Measurement of Smad3-directed Transcriptional Activity-We employed mink lung cells stably transfected with a minimal promoter from the plasminogen activator inhibitor-1 (PAI-1) gene linked to luciferase (30). Cells were also transiently transfected (31) with pSBE4-Luc (46), kindly provided by Dr. Bert Vogelstein, to measure Smad3 transactivation. After treatment with TGF␤ and AMPK activators for 24 h, cytosolic luciferase activity was measured as described (31).
Subcellular Fractionation-Cells were homogenized at 4°C in 0.25 M sucrose containing 10 mM Tris-Cl (pH 8.0), 1 mM KCl, 1.5 mM MgCl 2 , 1 mM dithiothreitol with protease inhibitors (Roche Applied Science). The cell lysate was centrifuged at 3000 ϫ g for 10 min at 4°C, and the resulting pellet and supernatant were collected as the nuclear and cytoplasmic (post nuclear supernatant) fractions.
Data Analysis and Statistics-Data are mean Ϯ S.D. from 3-4 independent experiments. Statistical significance was calculated by unpaired Student's t test or by analysis of variance with Bonferroni multiple correction using InStat (GraphPad) (asterisks represent p Ͻ 0.05 in the figures).
To directly determine whether AMPK blocks MFT, we added AMPK activators to cells exposed to TGF␤. We then measured induction of ␣SMA and the ED-A FN fibronectin splice variant, well established markers of the MFT phenotype. AICAR, adiponectin, and A-769662 blocked the 2.1-fold incre-ment in ␣SMA in cells exposed to TGF␤ (Fig. 3A). Similarly, the AMPK activators prevented induction of ED-A FN by TGF␤. Quiescent mesangial cells expressed minimal ED-A FN by immunocytochemistry and ELISA, but 48 h after TGF␤ abundant ED-A FN extracellular fibrils were detected (Fig. 3, B-F), similar to induction of ED-A FN in other cellular models of MFT (6). A-769662 blocked the 2.6-fold increase of ED-A FN by TGF␤ (Fig. 3, C, D, and F). AICAR and adiponectin also inhibited ED-A FN expression (Fig.  3F). Taken together, these results suggest that AMPK negatively regulates MFT.
Genetic Evidence That AMPK␣ Blocks TGF␤-stimulated MFT-To further investigate negative regulation of MFT by AMPK, we asked whether a dominant negative, kinase-dead AMPK␣ mutant that antagonizes AMPK signaling (40) prevents inhibition of MFT by AMPK activators. Adenovirus transduction with dnAMPK␣2 prevented inhibition of TGF␤-stimulated collagen I secretion by AICAR, adiponectin, and A-769662 (Fig. 4,  A and B). Transduction with an adenoviral vector expressing GFP had no effect on basal or TGF␤-stimulated collagen I secretion, demonstrating that the effect of dnAMPK␣2 was not a nonspecific consequence of adenoviral gene transfer. Moreover, AICAR blocked expression of collagen I in cells transduced with Ad.trGFP. In cells transduced with Ad.dnAMPK␣2, the 2.3-fold induction of ␣SMA by TGF␤ was not inhibited by AICAR, adiponectin and A-769662 (Fig.  4C). Similarly, AMPK activators did not block the 2.7-fold elevated secretion of ED-A FN in cells transduced with Ad.dnAMPK␣2 (Fig. 4D). Basal and stimulated levels of ␣SMA or ED-A FN were not altered by adenoviral transduction with Ad.trGFP. These results with dnAMPK␣2 support the hypothesis that AMPK signaling blocks MFT.

AMPK is required for inhibition of TGF␤-stimulated collagen I by A-769662 and AICAR.
Similarly, in AMPK ␣ Ϫ/Ϫ null fibroblasts A-769662 and AICAR did not inhibit expression of the MFT markers ␣SMA and ED-A FN. Induction of ␣SMA by TGF␤ was similar in AMPK ␣ ϩ/ϩ or AMPK ␣ Ϫ/Ϫ cells, demonstrating that the signals linking TGF␤ receptor activation to induction of ␣SMA are unaffected in the null fibroblasts (Fig. 5B). However, the 2.5fold induction of ␣SMA in AMPK ␣ ϩ/ϩ cells treated with TGF␤ was not inhibited by A-769662 or AICAR in AMPK ␣ Ϫ/Ϫ null fibroblasts (Fig. 5B). A-769662 and AICAR also failed to block ED-A FN secretion in AMPK ␣ Ϫ/Ϫ fibroblasts exposed to TGF␤ (Fig. 5C). Collectively, these experiments in AMPK ␣ Ϫ/Ϫ fibroblasts provide strong evidence that AMPK negatively regulates TGF␤-stimulated MFT.
We next asked whether AMPK activity is sufficient to block TGF␤-stimulated MFT. An AMPK␣2 mutant consisting of only the kinase domain and lacking the auto-inhibitory and ␤/␥ interaction domains (Fig. 6A) is constitutively active (caAMPK␣2), even in the absence of AMPK activators (26). We expressed caAMPK␣2 by means of an adenoviral gene transfer vector. Transduction of Ad.caAMPK␣2 had no effect on basal collagen I secretion, but it blocked TGF␤stimulated expression (Fig. 6B). Neither basal nor TGF␤stimulated collagen I secretion were affected by transduction with Ad.trGFP (Fig. 6B). caAMPK␣2 also blocked induction of ␣SMA and ED-A FN protein by TGF␤ (Fig. 6, C and D). Transduction with Ad.trGFP had no effect on induction of ␣SMA and ED-A FN. Taken together, these results with dn.AMPK␣2, AMPK ␣ Ϫ/Ϫ null fibroblasts, and ca.AMPK␣2 demonstrate that AMPK␣ is sufficient to prevent MFT in cells exposed to TGF␤.
AMPK Blocks Smad3-mediated Transcription without Inhibiting Smad3 Phosphorylation or Nuclear Translocation-To determine the mechanism underlying inhibition of MFT by AMPK, we studied signals downstream of TGF␤ receptor activation. Because Smad3 is essential for TGF␤-stimulated MFT (41-44), we asked whether AMPK blocks Smad3 phosphorylation or nuclear translocation.
Immunofluorescent analysis showed rapid TGF␤-stimulated COOH-terminal activating phosphorylation of Smad3 and nuclear localization. PSmad3 was nearly undetectable in the cytosol or nucleus of unstimulated cells, but after 30 min of TGF␤ 83% of cell nuclei were positive for PSmad3 (Fig. 7,  A and B). Two hours after TGF␤, nuclear levels of PSmad3 started to modestly decline. The AMPK activators AICAR and A-769622 did not reduce the staining intensity or nuclear localization of PSmad3 in response to TGF␤ (Fig. 7,  A and B). An unexpected finding was that AMPK activators alone transiently increased PSmad3 levels and translocation, albeit much less intensely than with TGF␤ alone or TGF␤ plus the AMPK activators (Fig. 7B). These results demonstrate that AMPK does not inhibit Smad3 phosphorylation at the activating epitope. Moreover, AMPK does not inhibit nuclear translocation of PSmad3.
Because AMPK did not block phosphorylation or nuclear translocation of Smad3, we investigated whether AMPK inhibited Smad3-mediated transcription downstream of the nuclear localization step. To that end we employed reporter cells with three Smad3-sensisitive cis-elements in the context of a heterologous human PAI-1 promoter driving transcription of luciferase (30,45). TGF␤ increased luciferase reporteractivity24.7-fold (Fig.8A),andtheincreaseinSmad3dependent transcription was blocked by a TGF␤-neutralizing antisera, demonstrating specificity for TGF␤ (data not shown). Both AICAR and A-769662 blocked the increase in luciferase activity in cells exposed to TGF␤ (Fig. 8A). Activation of Smad3-dependent transcription was not inhibited by transduction with Ad.trGFP but was blocked by the dominant negative AMPK␣2 mutant (Fig. 8A). We confirmed that AMPK blocked Samd3-dependent transcription in mesangial cells transfected with pSBE4-Luc (Fig. 8B), which is transactivated only by Smad3 and not by other R-Smad proteins (46). Collectively, these data suggest that AMPK inhibits Smad3-mediated transcription by a mechanism downstream of TGF␤-stimulated Smad3 phosphorylation and nuclear translocation.
We also asked whether active AMPK was present in mesangial cell nuclei, consistent with a role for AMPK in transcriptional regulation. Because it is well established that AMPK␣1 does not translocate to the nucleus (47, 50), we assessed translocation of AMPK␣2 in subcellular fractions. The AMPK␣2 subunit was barely detectable in nuclei of quiescent cells, but the kinase rapidly translocated from cytosol to the nucleus when cells were stimulated with AICAR ( Fig.  9). Nuclear translocation of AMPK␣2 correlated with the appearance of P172ThrAMPK␣ (active kinase) in the nucleus. AICAR-stimulated nuclear localization and activation of AMPK␣2 was unaffected by addition of TGF␤, and addition of TGF␤ alone weakly elevated P172ThrAMPK␣ at 1 h, which presumably reflected phosphorylation of AMPK␣ by the TGF␤-activated protein kinase (49,50). However, we note that the level of AMPK␣ activation by TGF␤ was of insufficient magnitude and duration to inhibit MFT. These results demonstrate that AMPK␣2 can translocate to the nucleus in an activated form in mesangial cells and support a role for AMPK␣ in regulation of Smad3-dependent transcription.

DISCUSSION
MFT, a metaplastic change in cell phenotype, is essential for the development of profibrotic effector cells in the wound healing response and is induced by extracellular cues such as growth factors, cytokines, and mechanical stress. Yet the molecular mechanisms that negatively regulate MFT are poorly understood. In this study, we demonstrate that AMPK is sufficient to repress TGF␤ induction MFT of mesangial cells in vitro.
AMPK as a Negative Regulator of MFT-In eukaryotes AMPK has an evolutionarily conserved role regulating the catabolic and anabolic metabolism of glucose and fatty acids at the cellular and whole body levels (10 -12). Recent studies have revealed novel functions for AMPK, including regulation of a cell cycle checkpoint (51), the decision to enter autophagy or apoptosis (52,53), and cell fate decisions in development (54). Our experiments show that two mechanistically different pharmacological activators of AMPK block induction of myofibroblast marker proteins in response to TGF␤. A caveat for interpreting results with AICAR is that 5-aminoimidazole-4-carboxamide-1-d-ribofuranosyl 5Ј-monophosphate can weakly stimulate glycogen phosphorylase (55,56) and inhibit fructose-1,6-bisphosphatase (57). Unlike AICAR, however, A-769662 does not affect glycogen phosphorylase or fructose-1,6-bisphosphatase activity (19). A-769662 does not alter the activity of a broad spectrum of proteins in a high-throughput screen (19), suggesting that A-769662 is selective for AMPK. The finding that AICAR and A-769662 failed to repress MFT marker proteins in AMPK ␣1,2 null fibroblasts treated with TGF␤ suggests that the pharmacological agonists were specific for AMPK. Regarding the role of MFT in fibrosis, an important phenotypic change in mesangial cells undergoing MFT is secretion of the fibrillar collagen type 1, which is not secreted by normal mesangial cells in vivo but is robustly induced in glomerulosclerosis. Excessive secretion of fibronectin and the basement membrane protein collagen IV is also observed in mesangial-myofibroblasts in glomerulosclerosis in vivo. It is therefore significant that the AMPK activators blocked induction of collagen I, collagen IV, and fibronectin in mesangial cells.
The notion that AMPK negatively regulates TGF␤-stimulated MFT was also demonstrated in this study using three genetic approaches. First, mutation of lysine 45 to arginine in the AMPK␣2 kinase domain results in a dominant negative mutant that antagonizes AMPK signaling (40). Adenoviral transduction of this AMPK␣2 mutant prevented repression of collagen I, ␣SMA, and ED-A FN in mesangial cells stimulated with TGF␤. Second, a constitutively active mutant of AMPK␣2, which stimulates AMPK signaling (26), prevented induction of the MFT marker proteins. These findings also suggest that AMPK activity is sufficient in mesangial cells to block the MFT transition in response to TGF␤ and concurs with our results with AICAR and A-769662. Finally, AMPK activators were unable to block MFT in murine embryonic fibroblasts null for AMPK␣ subunits. Double knockouts of AMPK␣1 and AMPK␣2 are embryonic lethal (22), thus mesangial cells lacking both AMPK␣ subunits could not be prepared. These results in murine embryonic fibroblasts are also important, because they suggest that the negative effect of AMPK on MFT is not restricted to glomerular mesangial cells and that AMPK might be a more general inhibitor of MFT in other cellular contexts. A Smad3-based Mechanism for Inhibition of MFT by AMPK-As a window into potential mechanisms that underlie negative regulation of MFT by AMPK, we asked whether AMPK in- Inhibition of Smad3-stimulated signaling by AMPK occurs downstream of nuclear translocation and involves transcriptional regulation. AMPK blocked TGF␤-stimulated transactivation of a luciferase reporter construct (pSBE4-Luc) that is highly selective for Smad3. AMPK also inhibited activation of the human PAI-1 promoter, an informative model of Smad3dependent transcription (45, 58 -60). Stimulation of the human PAI-1 promoter by TGF␤ has been previously shown to require Smad3 and three Smad3-binding cis-elements (45, 58 -60). In addition, PAI-1 gene induction is relevant to the study of MFT and glomerular fibrosis: several laboratories have documented that PAI-1 is potently induced in MFT and that it contributes to the increase in fibrosis that results from MFT (61)(62)(63). TGF␤ also induces transcription and expression of PAI-1 in human mesangial cells (64), and genetic depletion of PAI-1 augments mesangial matrix protein accumulation and glomerular fibrosis in an experimental model of chronic kidney disease (65). Our results demonstrate that AMPK prevents transcriptional activation of the PAI-1 promoter by TGF␤. Inhibition of Smad3-dependent activation of the PAI-1 promoter by the AMPK activators was antagonized by adenovirus transduction of the dnAMPK␣2 mutant, confirming that repression by AICAR and A-769662 required AMPK activity. Translocation of active AMPK␣2 to the mesangial cell nucleus, which has previously been demonstrated in other cells types (47,48), provides further support for the notion that AMPK regulates Smad3-directed gene expression at the level of transcription. Whether . Nuclear translocation of AMPK␣2 and localization of active AMPK␣. Quiescent mesangial cells were stimulated for 1 or 2 h with 0.5 mM AICAR, 0.5 ng/ml TGF␤, or AICAR plus TGF␤. Subcellular cytosolic and nuclear fractions were prepared by centrifugation and analyzed by Western blotting for the indicated proteins. The blots were reprobed with ␤-actin or lamin-A to confirm equal protein loading and to assess the relative purity of the cytosolic and nuclear fractions. Results are representative of three independent experiments.