Differential Effect of Activin A and BMP-7 on Myofibroblast Differentiation and the Role of the Smad Signaling Pathway

Abstract

purpose. To investigate myofibroblast differentiation and signal transduction induced by TGF-β family members activin A and bone morphogenetic protein (BMP)-7.

methods. Transcription of activin and receptors (ActR) for activin A and BMP-7 was detected by RT-PCR. Levels of marker proteins for differentiation and phosphorylation of similar to mothers against decapentaplegics (Smads) were quantified by Western blot analysis in response to BMP-7, activin A and follistatin. Transfection with antisense Smad2/3 was performed to evaluate signal transduction.

results. Activin A and receptors (ActR-I, ActR-IB, ActR-II) are transcribed in corneal fibroblasts. Compared with TGF-β1 or serum, activin A but not BMP-7 increased α-smooth muscle (SM) actin and actin-binding proteins such as SM myosin, α-actinin, and vinculin. Talin, paxillin, and desmin were not induced and vimentin was downregulated by activin. Activin also induced extracellular matrix proteins fibronectin and integrin β1. Activin-dependent accumulation of proteins was blocked by follistatin. Regarding signal transduction, activin A induced phosphorylation of Smad 2, and BMP-7 induced Smad 1, both of which were inhibited by follistatin. Transfection with antisense Smad 2/3 prevented activin-induced expression and accumulation of α-SM actin.

conclusions. TGF-β proteins have different functions in the cornea. Activin A and TGF-β1, but not BMP-7, are regulators of corneal keratocyte differentiation and may play a role during myofibroblast transdifferentiation. Smad 2/3 signal transduction seems to be important in the regulation of muscle-specific genes. Further investigation of Smad signaling may help to better understand the function of TGF-β family members in the cornea.

Cellular proliferation, differentiation, and wound healing in the cornea are largely regulated by polypeptide growth factors. Transforming growth factor (TGF)-β is unique among the cytokines in the cornea, because it simultaneously inhibits epithelial proliferation and stimulates keratocyte proliferation and migration. ^{1 2 3} Prototypic members of the TGF-β superfamily, such as TGF-β1 and -β2 also regulate keratocyte differentiation ^{4 5 6 7} . Although their role in the cornea is relatively well investigated, the functional significance of most other members of the TGF-β family in the cornea is unknown. 

Approximately 30 related dimeric proteins belong to the TGF-β superfamily. ^{8 9} Based on structural homology, at least three major groups of proteins can be differentiated: TGF-βs, bone morphogenetic proteins (BMPs) and activins. The biological effects of these cytokines are mediated by signaling through two families of transmembranous serine-threonine kinase receptors. ^{9 10 11} After ligand binding, a heteromeric complex is formed by a type I and a type II receptor that initiates phosphorylation of the type I receptor and activation of downstream signaling cascades. ⁹ According to the different groups of TGF-β, three classes of receptors have been described: TGF-β receptors (TGF-βRs), BMP-receptors (BMPRs) and activin receptors (ActRs). It is well established that both the corneal epithelium and stroma express TGF-β receptors and that their ligands induce a variety of cellular functions. ^{12 13} Also, BMP-receptors and several BMPs have been described in human corneal epithelium and stroma. ^{14 15} Signaling through BMPRs has been shown to induce inflammatory mediators such as NF-κB and to modulate apoptosis. ¹⁴ Furthermore, ligands to BMPR have an effect on fibroblast chemotaxis. ¹⁶ It is presently unknown whether the third receptor system, ActR, is expressed in the human cornea and what function their ligands might have. 

After receptor binding, phosphorylation of the type I receptor initiates signal transduction through proteins belonging to the“ similar to mothers against decapentaplegic” (Smad) family. ^{17 18} On the basis of their function receptor, activated Smad proteins can be differentiated into two groups serving different types of receptors. ^{17 18} Activation of one arm of the Smad signaling pathway is induced by ActR-IB and results in phosphorylation and subsequent nuclear translocation of Smads 2 and 3. The other arm of the Smad signaling pathway is induced by ActR-I and results in activation of Smads 1 and 5. Therefore, binding of TGF-β family members to ActRs can initiate signaling through two different Smad pathways. This is in contrast to BMPR, which can activate only Smads 1 and 5 and to TGF-βR, which can activate Smads 2 and 3. Nuclear translocation of Smad 2 in response to TGF-β1 has recently been shown in canine keratocytes. ¹⁹ It is interesting that this process seems to depend on cell density, which can modulate myofibroblast differentiation. ^{5 19} The process of myofibroblast transdifferentiation is characterized by the expression of muscle-specific genes such as α-actin. ²⁰ In the cornea, myofibroblast transdifferentiation is instrumental for the induction of scar formation and the reduction of corneal transparency. ^{7 21 22 23}  

In the present study we show that activin and several ActR are transcribed in corneal fibroblasts. To learn about the biological significance of the ActR system in the cornea, we have investigated two prototype ligands (activin A and BMP-7) regarding their effect on fibroblast differentiation. Activin A and TGF-β 1 had similar effects on proteins related to myofibroblast differentiation, migration, and cell adhesion. In contrast, the effect of BMP-7 was less pronounced. Follistatin, a protein that blocks activin receptors by binding activin A or BMP-7, neutralized the effect of activin A and BMP-7 on activation of intracellular signals and induction of gene expression. ²⁴ Because these results suggest a functional importance of Smads 2 and 3 for activin-dependent keratocyte differentiation, we performed transfection with an antisense DNA construct of Smad 2/3. Interruption of this signaling cascade also blocked the effect of activin A on α-SM actin transcription, thus confirming that the expression of muscle-specific proteins in corneal fibroblasts is in part regulated by Smad signals. 

Materials and Methods

Materials

A total RNA isolation kit (RNAgents), an mRNA purification system (polyATtract), RNase-free DNase, and Moloney murine leukemia virus reverse transcriptase (M-MLV RT) were obtained from Promega (Madison, WI). Thermus aquaticus (Taq) DNA Polymerase was obtained from Qiagen (Hilden, Germany). Digoxygenin (DIG) probe synthesis mix, DIG buffer (EasyHyb), a DIG washing set, a protein G immunoprecipitation kit, protease inhibitor (Complete), and bFGF (FGF-2) were from Roche Diagnostics (Mannheim, Germany). Enhanced chemiluminescence (ECL) Western blot analysis system and nitrocellulose membranes came from Amersham (Buckinghamshire, UK). 10% SDS-3-(N-morpholino)propanesulfonic acid (MOPS) gels (NuPAGE Bis-tris) or 3% to 8% Tris-acetate gels (NuPAGE) were obtained from Invitrogen (San Diego, CA). Antibodies against α-SM actin, SM myosin (light- and heavy-chain), α-actinin, vinculin, talin, paxillin, integrin β1, fibronectin, desmin, and anti-mouse IgG conjugated to FITC and keratinocyte basal medium were purchased from Sigma (St. Louis, MO) and antibody against vimentin from Dako (Glostrup, Denmark). Polyclonal antibodies against Smads 1 and 2 were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Polyclonal antibody against phospho-Smad 1 (Ser463/465) and polyclonal antibody against phospho-Smad 2 (Ser465/467) were from Upstate Biotechnology (Lake Placid, NY). Dulbecco’s modified Eagle’s medium (DMEM) came from Gibco (Grand Island, NY), and TGF-β1 and follistatin were from R&D (Minneapolis, MN). Recombinant activin A was a generous gift from Yuzuru Eto (Ajinomoto Co., Inc., Kawasaki, Japan). Recombinant BMP-7 was a generous gift from Donald Jin (Creative Biomolecules, Hopkinton, MA). 

Cell Culture

Human corneas stored for less than 24 hours in preservation medium (Likorol; Chauvin-Opsia, Labege, France) at 4°C were used to initiate keratocyte outgrowth cultures in DMEM with 10% fetal bovine serum (FBS), as described. ¹⁵ For experiments, cells from passages 2 to 4 were used. Because seeding density is crucial in relation to the status of differentiation ^{5 19} an intermediate density (150 cells/mm²) was used that has been shown to allow the expression of differentiation-related proteins such as α-SM actin’s response to TGF-β. ¹⁹  

RNA Isolation and Reverse Transcription–Polymerase Chain Reaction

Total RNA and mRNA was isolated from fresh, snap-frozen human tissue samples or cultured corneal fibroblasts with a total RNA isolation kit (RNAgents) and an mRNA purification kit (polyATtract; both from Promega), as described. ¹⁵ To minimize the risk of contamination by genomic DNA the mRNA samples were digested by RNase-free DNase followed by phenol-chloroform extraction and isopropanol precipitation. 

The following primers were used: activin βA: sense, CAGGATGCCCTTGCTTTGGCTGAGA, and antisense, CCCACATGAAGCTTTCTGATCGCGT (282 bp, GenBank accession no. M13436, J03634); ActR-I: sense, GTACAATGGTAGATGGAGTGATGAT, and antisense, CATACCTGCCTTTCCCGACACAC (663 bp, GenBank L02911, Z22534); ActR-IB: sense, ACGTGTGAGACAGATGGGGCCTGC, and antisense, GCGATGATGCCTACCAGCTCCACCG (263 bp, GenBank U14722, Z22536); ActR-II: sense, ATCTAGCGAGAACTTCCTCCG, and antisense, GCCCTCACAGCAACAAAAATATAC (364 bp, GenBank D31770, M93415, X62381). (GenBank is provided by the National Center for Biotechnology Information, Bethesda, MD, and is available in the public domain at http://www.ncbi.nlm.nih.gov/genbank/) 

First-strand cDNA was synthesized by M-MLV RT at 42°C. PCR was performed using 1 ng single-strand cDNA with 3 U Taq DNA polymerase in a volume of 50 μL. After predenaturation at 95°C for 3 minutes, 35 cycles were performed including denaturation at 94°C for 1 minute, annealing at 55°C for 1 minute and extension at 72°C for 1 minute followed by 2% agarose gel electrophoresis. All the PCR fragments were cloned and sequenced. 

Western Blot Analysis

To investigate the effect of activin A and BMP-7 on keratocyte differentiation, Western blot analysis were performed. Fibroblasts were incubated in serum-free DMEM, with or without recombinant human activin A (100 ng/mL), ^{25 26 27} BMP-7 (200 ng/mL), ²⁸ TGF-β1 (1 ng/mL), FGF-2 (bFGF; 100 ng/mL) or 10% FBS for 3 days and solubilized in lysis buffer containing dithiothreitol and protease inhibitors. Total protein (80 μg per lane) was fractionated by 10% SDS-MOPS gels or 3 to 8% Tris-acetate gel and blotted onto nitrocellulose membranes. Membranes were then incubated with monoclonal antibodies against α-SM actin, SM myosin (heavy- and light-chain),α -actinin, vinculin, talin, paxillin, integrin β1, fibronectin, desmin, and vimentin and visualized with the ECL Western blot analysis system. The relative intensities of the protein bands were quantified using NIH Image software, ver. 1.62 (NIH Image is provided in the public domain by the National Institutes of Health, Bethesda, MD, and is available at http://www.nb.nih.ncbi.gov). Results are expressed as a percentage of the signal obtained from serum-free control cultures set as 100% (y-axis, Figs. 2 3 4 5 6 ). 

To investigate the effect of activin A and BMP-7 on time-dependent phosphorylation of Smads 1 and 2, fibroblasts were starved in serum-free DMEM for 1 day and incubated in serum-free DMEM without additives or with recombinant human BMP-7 (200 ng/mL) or activin A (100 ng/mL) for 5, 20, 35, and 50 minutes. Cell matrix proteins were extracted in the presence of dithiothreitol, sodium pyrophosphate, sodium orthovanadate, and a protease inhibitor (1 tablet/30 mL buffer; Complete; Roche). Western blot analyses were performed as described earlier with a polyclonal antibody against phospho-Smad 1(for BMP-7 induction) and phospho-Smad 2 (for activin induction) and polyclonal antibodies against total Smad 1 or 2 (phosphorylated and nonphosphorylated Smad). 

To investigate the effect of follistatin on activin-induced Smad activation and keratocyte differentiation, fibroblasts were incubated with 200 ng/mL follistatin for 30 minutes followed by stimulation with activin (100 ng/mL) or BMP-7 (200 ng/mL) for 30 minutes or 3 days and total protein isolation for Western blot. 

Immunoprecipitation

Cultured fibroblasts were incubated in serum-free DMEM without additives or with recombinant human activin A (100 ng/mL), TGF-β1 (1 ng/mL), FGF-2 (100 ng/mL) or 10%FBS for 3 days and solubilized in lysis buffer containing 50 mM Tris₂Cl (pH 8.0), 150 mM NaCl, 0.02% sodium azide, 100 μg/mL phenylmethylsulfonyl fluoride, 1 mM Na₃VO₄, 1% Triton X-100, and protease inhibitor (1 tablet/30 mL buffer, Complete; Roche). The lysate was then pelleted by brief centrifugation at 12,000 rpm. Protein concentrations in the supernatant were determined by Bradford assay. Protein (50 μg) in 500 μL lysis buffer was incubated with 5 μg monoclonal antibody against SM myosin (light chain) for 1 hour at 4°C, and 30 μL protein-G agarose (Roche) was used for protein absorption overnight according to the instructions of the manufacturer. The agarose protein complex was then pelleted by centrifugation at 12,000 rpm. Bound proteins from protein-G agarose were solubilized in SDS gel loading buffer. After heating to 100°C for 5 minutes and brief centrifugation at 12,000 rpm, samples were subjected to Western blot analysis. 

Construction of an Antisense Smad2/3 Expression Vector and Gene Transfection

To investigate the importance of the Smad signaling pathway for the induction of keratocyte differentiation by activin A, we tested the effect of inhibiting Smad2/3 signaling on expression and accumulation of α-SM actin protein. An antisense fragment of Smad2 (−1 to +199 of the coding sequence) was cloned into the HindIII-BamHI cloning site of the pEGFP-N3 vector (Clontech Laboratories, Palo Alto, CA) to overexpress antisense Smad transcripts fused with enhanced green fluorescent protein (EGFP). Constructs were confirmed by sequencing. Cultured fibroblasts were transfected with pEGFP-anti-Smad or pEGFP (as a control) using reagents (Lipofectamine and Plus; Gibco) for 3 hours in serum-free keratinocyte basal medium. To ensure that only transfected cells were investigated, we performed a selection for 3 weeks, during which transfected cells grew in serum-containing DMEM with 150 μg/mL geneticin (G418; Amersham) and nontransfected cells did not survive. After selection, cells were incubated in serum-free DMEM, with or without recombinant human activin A (100 ng/mL), for 3 days and were then solubilized for Western blot analysis. 

Northern Blot Analysis

To further confirm the effect of pEGFP-anti-Smad transfection on transcription of the α-SM actin gene, Northern blot analysis was performed. After transfection and 3 weeks of selection in medium containing genetticin, 1 × 10⁷ corneal fibroblasts expressing either EGFP-anti-Smad or EGFP mRNA were cultured in serum-free DMEM without additives or with recombinant human activin A (100 ng/mL) for 1.5 hours followed by lysis with a total RNA isolation kit (RNAgents; Promega). Total RNA (50 μg) from each sample was separated on formaldehyde agarose gels (1%) and subsequently blotted onto nylon membranes. The RNA blots were then hybridized in 42°C overnight with either α-SM actin or a GFP cDNA probe labeled with digoxygenin-dUTP and subjected to DIG-detection and ECL film exposure. 

In general, all experiments were performed in triplicate. 

Results

Presence of Activin A and Corresponding Receptors Transcripts in the Cornea

A cDNA fragment specific for the activin βA chain was amplified by RT-PCR from ex vivo corneal stroma (Act (ex vivo), 282 bp) and cultured corneal fibroblasts (Act, (ex vitro), 282 bp; Fig. 1A ). The cellular responses induced by activin A and BMP-7 were mediated by type I and type II ActR and BMPR. cDNA fragments specific for ActR-I (663 bp), ActR-IB (263 bp), and ActR-II (364 bp) were amplified from ex vivo corneal stroma (ex vivo) and cultured corneal fibroblasts (ex vitro; Fig. 1B ). 

Induction of Myofibroblast Differentiation in Stromal Fibroblasts

We compared the effect of activin A (Fig. 2 , Act) on keratocyte differentiation with that of TGF-β1 (TGF), because the latter had been thoroughly studied before. We also investigated two other mediators of corneal differentiation, FGF-2 (FGF) and serum (ser) as well as serum-free medium (co). Because we wanted to observe the maximal effect of FGF-2 we used a concentration of 100 ng/mL. 

The first set of experiments shows that activin A induced the expression of proteins that are related to myofibroblast differentiation or alteration of cell shape in relation to the extracellular matrix. Three of the most important proteins in this respect are α-SM actin and SM myosin heavy and light chains, which are markers for corneal myofibroblast transdifferentiation. ^{6 7} Figure 2 shows that both TGF-β1 and activin A increase levels of α-SM actin (42 kDa; Fig. 2A ) as well as SM myosin heavy chain (200 kDa; Fig. 2B ). Also SM myosin light chain (20 kDa; which was investigated by immunoprecipitation and Western blot) was strongly induced by both TGF-β1 and activin A compared with the control (Fig. 2C) . 

The second set of experiments shows that activin differentially induced the expression of other proteins related to myofibroblast differentiation. α-Actinin, an actin-bundling protein, as well as vinculin, talin, and paxillin mediate the link between actin filaments and extracellular matrix. In Figure 2 , we show that α-actinin (100 kDa; Fig. 2D ) and vinculin (116 kDa; Fig. 2E ) were strongly induced by activin A, whereas serum and TGF-β showed differential effects. In contrast activin A had no effect on the level of talin (225 kDa; Fig. 2F ) and paxillin (68 kDa; Fig. 2G ) which were both upregulated by serum. 

In a third set of experiments we show that activin A had a different effect on cytoskeleton proteins that are not directly related to myofibroblast differentiation. The level of vimentin (57 kDa) was reduced by activin A (Fig. 3A) , whereas the level of desmin (53 kDa) remained the same (Fig. 3B) . 

In a fourth set of experiments, activin also upregulated extracellular matrix proteins (fibronectin and integrin) that are essential for cell attachment. Levels of fibronectin (240 kDa) were induced by TGF-β1, activin, and serum but inhibited by FGF-2 (Fig. 4A) . Similarly, integrin β1 (120 kDa) was strongly augmented by TGF-β1, activin A, and serum (Fig. 4B) . 

Follistatin’s Inhibition of the Effect of Activin on Keratocyte Differentiation

To confirm that the effect of activin A on keratocyte differentiation is mediated by binding to ActR, we investigated the effect of follistatin. Follistatin binds to activin A and specifically blocks interaction of ActR with its ligand. ²⁴ Data shown in Figure 5 confirm that activin A increased α-SM actin. Pretreatment with follistatin inhibited this increase (Fig. 5) . Furthermore, activin A decreased vimentin levels, and pretreatment with follistatin prevented this decrease (Fig. 5) . Therefore, blockage of ActR-activin interaction by follistatin inhibited the effect of activin A on keratocyte differentiation. 

Effect of BMP-7 on Keratocyte Differentiation

The effect of BMP-7 on keratocyte differentiation was studied in comparison with that of TGF-β, FGF-2, and serum. Data presented in Figure 6A show that the level of α-SM actin remained unchanged in the presence of BMP-7 (BMP) and FGF-2 (FGF) but was significantly induced by TGF-β1 (TGF) or serum (ser) in comparison with serum-free medium (co). SM myosin heavy chain remained unchanged by BMP-7 or FGF-2, but was significantly induced by serum (Fig. 6B) . 

Phosphorylation of Smads 1 and 2 by BMP-7 and Activin A

To further investigate the signal transduction pathway induced by BMP-7 or activin A, we studied phosphorylation of Smad 1 or 2 by activin receptor serine-threonine kinases, which is a critical step for the initiation of Smad-mediated transcriptional responses. Data presented in Figure 7 show that BMP-7 (BMP) time dependently induced phosphorylation of Smad 1 and that activin A (Act) time dependently induced phosphorylation of Smad 2. The level of phosphorylated Smads increased up to 35 minutes and decreased at 50 minutes after addition of BMP-7 or activin A. During the entire observation period, levels of total Smad 1 (65 kDa) or 2 (52 kDa) remained the same (Fig. 7) . 

Inhibition by Follistatin of the Phosphorylation of Smads by BMP-7 or Activin

To test the specificity of the effect of BMP-7 and activin A on Smad phosphorylation, we investigated the effect of follistatin. Phosphorylation of Smad 1 induced by BMP-7 was inhibited by follistatin as shown in Figure 8 . Similarly, phosphorylation of Smad 2 induced by activin A was inhibited by follistatin. The intracellular level of total Smad 1 and 2 proteins remained the same under each condition. In conjunction with the findings shown in Figure 5 , these data suggest the possibility that corneal fibroblasts contain a regulatory system composed of BMP-7 and activin A as inducer and follistatin as inhibitor of Smad-signaling activation through activin receptors. 

Effect of Antisense Smad Transfection on Activin-Induced Myofibroblast Transdifferentiation

To further confirm the importance of the Smad 2/3 signaling pathway for mediating activin-induced keratocyte differentiation, we selectively inhibited this pathway by antisense gene transfection. For this purpose we used two vectors. One contains a sequence encoding for the EGFP. The other contains EGFP fused with an antisense sequence corresponding to the human Smad 2 gene, which therefore blocks the expression of Smad 2 protein (EGFP-AntiSmad, Fig. 9A ). Due to the high similarity between Smads 2 and 3 the expression of Smad 3 is most likely also involved. After transfection and selection (for 3 weeks) the morphology of EGFP cells in serum-containing medium was almost identical with nontransfected cells, whereas EGFP AntiSmad cells were slim, with predominantly narrow cytoplasm (Fig. 9B) . This morphologic change may be related to the decreased expression of muscle specific proteins regulated by the Smad pathway, as described earlier. Preliminary experiments had shown that more than 60% of cultured fibroblasts were transfected before selection (data not shown). After 3 weeks of selective culture, nontransfected cells were eliminated. 

All transfected cells expressed mRNA specific for EGFP (Fig. 10) , which is not present in nontransfected cells. The level of EGFP expression was identical in all transfected cells and independent of the presence of activin A (Act) in the culture medium. Cells transfected only with EGFP expressed Smad 2/3 protein and therefore contained normal levels of Smad 2 protein, which is independent of stimulation of activin A (Fig. 10) . In contrast, transfection with EGFP anti-Smad impaired transcription of Smad 2/3, and therefore cells contained significantly less Smad 2 protein than cells transfected with EGFP (Fig. 10) . The level of Smad 2 protein was not changed by activin A. Similar experiments also indicated that expression of Smad 3 protein was inhibited in cells transfected with EGFP anti-Smad (data not shown). In concordance with our results obtained by Western blot analysis (Fig. 4) , activin A significantly induced α-SM actin in cells that express Smad 2/3 proteins (EGFP; Fig. 10 ). In contrast, activin had no effect on expression of α-SM actin in fibroblasts that contained low levels of Smad 2/3 due to transfection with the antisense Smad2/3 construct (EGFP-AntiSmad; Fig. 10 ). These results confirm that the expression of α-SM actin in fibroblasts is modulated by activin A through the Smad pathway. 

Discussion

Within the TGF-β superfamily activins represent a distinct group of molecules that can comprise one α and three β chains (βA,β B, and βC). ^{29 30} Depending on the arrangement of these chains several dimeric molecules can be formed. Dimers composed of two β chains are called activins, and dimers consisting of one α and one β chain are called inhibins. Here we show transcription of genes encoding for the βA chain as well as several activin receptors in corneal fibroblasts. As a prototypic molecule, we have investigated the function of the bioactive molecule activin A that consists of two monomeric βA chains linked by disulfide bonds. ³⁰ Activin A as well as follistatin belong to a regulatory system in which follistatin inhibits receptor binding of activins-inhibins and therefore downregulates their biological effects. ^{24 30} Such a system functions, for example, in the ovary or the prostate. ^{31 32} In cultured corneal fibroblasts, follistatin blocks the effect of activin on protein expression and signal transduction. This suggests the possibility that the activin-follistatin system may also be present in the cornea and could be involved in the regulation of cellular functions. 

The major function of activin A and BMP-7 is the regulation of cell differentiation. During embryogenesis, these molecules are instrumental for axis development and organogenesis in a variety of species. ³³ In adults, activins function as hormone-type feedback regulators in the reproductive system. They circulate in the vasculature to regulate the release of follicle-stimulating hormone. ³⁴ Activin A controls several aspects of hematopoiesis ^{25 26 27} and regulates cell differentiation in the ovary, placenta, prostate, and testis. ^{30 35} Furthermore, the presence of activin has be related to wound healing. During cutaneous wounding, activin is released from the vasculature as well as from activated monocytes and macrophages. ³⁶ Wound fluid collected after severe burns induces the differentiation of erythroid cells, an effect that is most likely due to activin A. ³⁷ Finally, activin A is present in skin wounds, and its expression in dermal keratinocytes is increased after wounding. ^{37 38} In the current study, activin A modulated the levels of several differentiation-related proteins in corneal fibroblasts. Among the proteins under investigation the expression ofα -SM actin and SM-myosin was most prominently affected. Both of these proteins are normally present in muscle cells, and their expression in fibroblasts is indicative of a transition into an activated phenotype, which is called myofibroblast transdifferentiation. In the corneaα -SM actin is induced in activated keratocytes within the wound but not in neighboring cells and serves as a marker for myofibroblast differentiation. ^{22 23 39} This process has been shown to be influenced by activin A in nonocular tissues. In pulmonary fibrosis, scar formation is due to the activity of myofibroblasts, and activin A has been observed in remodeling lesions associated with interstitial pulmonary fibrosis. ⁴⁰ Furthermore, activin A increases the number of fetal lung fibroblasts immunopositive for α-SM actin. ⁴¹ Similarly, activin modulates the growth of vascular smooth muscle cells that also express SM actin and myosin. ⁴²  

In several studies, investigators have looked at the regulation ofα -SM actin or SM-myosin in a variety of nonocular tissues, as well as the cornea. It has been shown that wound-related cytokines such as platelet-derived growth factor (PDGF) and FGF-2 inhibit the expression of α-SM actin or SM-myosin, which is supported by our results. ^{6 43 44} In contrast, TGF-β1 has been shown to upregulate the expression of these proteins and to govern various aspects of myofibroblast differentiation in the context of wound healing. ^{6 7} Herein, we present evidence that a second TGF-β family member, activin A (but not BMP-7) has similar, yet not identical functions as TGF-β1 on keratocyte differentiation. It is notable that a similar common action was observed in RPE cells where both activin A and TGF-β seem to attenuate the effect of promitogenic cytokines in the context of proliferative vitreoretinopathy. ^{45 46}  

The mechanism by which activin modulates the expression of muscle-specific genes are unknown. For TGF-β a direct effect on the target gene has been shown and it seems therefore possible that activin A follows a similar mechanism. Investigations of the effect of TGF-β1 on α-SM actin gene expression have shown that the first 125 bp of theα -SM actin promoter are sufficient to confer TGF-β responsiveness. ⁴⁷ Within this region a TGF-β control element was identified that can also be found on the myosin promoter. Further studies are needed to identify response elements that can be induced by activin A. In addition to a direct mechanism, the effect of activin A could be indirect—for example, based on interactions with other cytokines. Corneal fibroblasts can express a variety of growth factors including TGF-β family members such as TGF-β1, β2, andβ 3 and BMP-2, -4, -5 and -7. ^{14 15 48} Because there are no data concerning the induction of cytokine gene expression by activin A during postnatal life, it should be determined whether activin A can induce expression of cytokines that modulate myofibroblast transdifferentiation. Activin gene expression can be induced by factors that are important in wound healing. For example, PDGF induces expression of activin A in bone marrow cells. ⁴⁹ Furthermore, TGF-β1 induces expression of activin A in differentiated cell lines, suggesting that TGF-β1 can stimulate the secretion of activin A. ⁵⁰ This may imply that the modulating effects of TGF-β1 and PDGF on myofibroblast transdifferentiation could be partly due to an induction of activin. This would also explain why some of the effects of activin A and TGF-β1 are similar. 

The effect of other members of the TGF-β superfamily on gene expression during differentiation seems to differ from that of activin A and TGF-β1. Activin A and BMP-7 could not induce mesenchymal transdifferentiation in NMuMG breast epithelial cells in comparison with significant induction by TGF-β. ⁵¹ BMP-12 and -13 inhibit terminal differentiation of myofibroblasts by inhibiting the expression of myosin. ⁵² Similarly, BMP-2 downregulates myosin and simultaneously induces markers for osteoblast differentiation. ⁵³ In the light of the high structural homology of the TGF-β family members, as well as overlapping affinity with transmembranous receptors, the diversity of the effect of members of this family on cellular differentiation is surprising. During the recent past, signal-transduction pathways have been investigated, and these studies allow insight in the divergent modes of action of various TGF-β family members. Although TGF-β and activin A bind to different receptors, they use the same signal-transduction pathway consisting of Smads 2, 3, and 4. ⁵⁴ In response to activin A or TGF-β, the carboxyl-terminal domains of Smads 2 and 3 are essential for phosphorylation of Smad 2/3, association with Smad 4, translocation into the nucleus, and transcriptional response. ^{55 56} To initiate transcription of target genes, Smads have to interact with transcription factors. Smad 2 has been shown to interact with FAST-1 and Smads 3 and 4 interact with Ap 1, ATF2 or Sp 1. ^{56 57 58 59} Furthermore, specific target genes for Smad 2 have been identified, such as ARE(Mix.2 promoter) or TARE(gscd promoter). ^{55 60}  

The finding that BMP-7 causes transcriptional responses that differ from those evoked by activin A can also be explained by the nature of the signal-transduction cascade. BMP-7 has a specific binding affinity to ActR-I, which activates Smads 1 and 5. ⁶¹ Consequently, transcription factors induced by BMP-7 are also different. Target DNA-binding proteins of Smad 1 specific for BMP-induced gene transcription are Hoxc-8, STAT, and OAZ. ^{62 63 64} One of the genes that are directly controlled by Smads 1 and 5 is the homeobox gene Tlx2. ⁶¹ Of note, BMP-7 induces transcription of genes specific for α-SM actin and SM myosin heavy chain during myofibroblast transdifferentiation in vascular smooth muscle cells. ⁶⁵ This is in contrast to our findings in corneal fibroblasts, where transcription of α-SM actin is regulated by activin-Smad 2/3 signaling and is independent of BMP-7. This suggests the presence of unknown mechanisms that control a cell type–dependent regulation of TGF-β signaling. Furthermore, it is currently not known how TGF-β family members, such as activin A and TGF-β1, can have very divergent functions, although their intracellular signals are mediated by identical Smad proteins. This could be due in part to the modulating effect of transcriptional repressors or activators that can modulate the activity of Smad proteins. Coactivators, such as C BP/P300 or corepressors such as Ski or SnoN, can regulate Smad-dependent transcriptional activity by binding to the MH2 domain of Smads 3 and 4 ^{66 67 68} . C BP, Ski, and SnoN play important roles regarding the regulation of transcription of numerous genes in vertebrates as well as Drosophila. Differential expression or activation of these regulators in association with Smad proteins may explain the diversity of the biological effects induced by TGF-β, activin, and BMPs. ⁶⁹ Another possible explanation is that other signal transduction pathways are used. It has been shown that TGF-β induces fibronectin synthesis through a c-Jun N-terminal kinase–dependent, Smad4-independent pathway. ⁷⁰ Whether activin A also uses other pathways remains to be investigated. Further analysis of factors that modulate the Smad signaling pathway should help to better understand the function of TGF-β family members in the cornea. 

 Supported by Deutsche Forschungsgemeinschaft (DFG), Bonn, Germany (Kr 993/12-1) and Gertrud Kusen-Stiftung Hamburg.

 Submitted for publication February 26, 2001; revised July 6, 2001; accepted July 24, 2001.

 Commercial relationships policy: N.

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Activin A and ActRs are transcribed in ex vivo corneal stroma (ex vivo) and cultured cornea (ex vitro). (A) Activin βA chain (282 bp; Act); (B) ActR-I (663 bp), ActR-IB (263 bp), and ActR-II (364 bp). One representative experiment is shown.

Figure 2.

 

Effect of activin A (100 ng/mL; Act) on intracellular levels of markers for myofibroblast differentiation and adhesion in cultured fibroblasts compared with the effect of TGF-β1 (1 ng/mL; TGF), FGF-2 (100 ng/mL; FGF), and serum (10%; ser). One representative Western blot for (A) α-SM actin (42 kDa), (B) SM myosin heavy chain (200 kDa), (C) SM myosin light chain (20 kDa), (D) α-actinin (100 kDa), (E) vinculin (116 kDa), (F) talin (225 kDa), and (G) paxillin (68 kDa) is shown with quantification (control medium set as 100%).

Figure 3.

 

Effect of activin A (100 ng/mL; Act) on intracellular levels of intermediate filaments in cultured fibroblasts compared with the effect of TGF-β1 (1 ng/mL; TGF), FGF-2 (100 ng/mL; FGF), and serum (10%; ser). One representative Western blot for (A) vimentin (57 kDa) and (B) desmin (53 kDa) is shown with quantification (control medium set as 100%).

Figure 4.

 

Effect of activin A (100 ng/mL; Act) on extracellular levels of matrix proteins in cultured fibroblasts compared with the effect of TGF-β1 (1 ng/mL; TGF), FGF-2 (100 ng/mL; FGF), and serum (10%; ser). One representative Western blot for (A) fibronectin (240 kDa) and (B) integrin β1 (120 kDa) is shown with quantification (control medium set as 100%).

Figure 5.

 

Follistatin (200 ng/mL) inhibits the effect of activin A (100 ng/mL) on a marker for myofibroblast differentiation and an intermediate filament in cultured fibroblasts. One representative Western blot for α-SM actin (42 kDa) and vimentin (57 kDa) is shown.

Figure 6.

 

Effect of BMP-7 (200 ng/mL; BMP) on intracellular levels of markers for myofibroblast differentiation in cultured fibroblasts compared with the effect of TGF-β1 (1 ng/mL; TGF), FGF-2 (100 ng/mL; FGF), and serum (10%; ser). One representative Western blot for (A) α-SM actin (42 kDa) and (B) SM myosin heavy chain (200 kDa) is shown with quantification (control medium set as 100%).

Figure 7.

 

BMP-7 (BMP; 200 ng/mL) induced time-dependent phosphorylation of Smad 1 (top), and activin A (Act; 100 ng/mL) induced time-dependent phosphorylation of Smad 2 (bottom). This representative Western blot analysis shows levels of phosphorylated Smad 1 or 2 proteins in comparison with total Smad 1 or 2 proteins after incubation with BMP-7 or activin A for between 5 and 50 minutes (m).

Figure 8.

 

Follistatin (200 ng/mL) inhibits phosphorylation of Smad 1 by BMP-7 (200 ng/mL; top) and phosphorylation of Smad 2 by activin A (100 ng/mL; bottom). This representative Western blot analysis shows levels of phosphorylated Smad 1 or 2 proteins in comparison with total Smad 1 or 2 proteins after incubation with BMP-7 or activin A alone as well as with BMP-7 or activin A after pretreatment with follistatin.

Figure 9.

 

Transfection of fibroblasts with an antisense cDNA fragment specific for Smad 2. (A) Map of pEGFP-antisense construct containing the empty multiple cloning site (MCS) of pEGFP (lightly shaded bar) and the antisense cDNA fragment of Smad 2 (darkly shaded bar). (B) Phase-contrast micrograph shows the morphology of pEGFP (EGFP) and pEGFP-antisense transfected (EGFPAntiSmad) corneal fibroblasts.

Figure 10.

 

Activin A induces expression of α-SM actin through Smad 2. First panel: representative Northern blot shows transcription of GFP in cells transfected with pEGFP (EGFP) or with pEGFP-AntiSmad (EGFPAntiSmad) grown in the absence (co) or presence of activin A (100 ng/mL; Act). Second panel: a representative Western blot shows expression of Smad 2 protein in cells transfected with pEGFP or with pEGFP-AntiSmad. Third panel: representative Northern blot analysis for α-SM actin mRNA. Fourth panel: representative Western blot forα -SM actin protein.

The authors thank Brigitte Erber for experienced help concerning cell culture and immunostaining, Klaus Rohrschneider for statistical analysis, and Bernhard Mechlers’ laboratory (German Cancer Research Center, Heidelberg, Germany) for help with Northern blot analysis and computer imaging.