Activation of Transforming Growth Factor-β1/p38/Smad3 Signaling in Stromal Cells Requires Reactive Oxygen Species–Mediated MMP-2 Activity During Bone Marrow Damage


  • Lin Wang,

    1. Department of Pediatrics, West Virginia University, School of Medicine, Morgantown, West Virginia, USA
    Search for more papers by this author
  • Suzanne Clutter,

    1. Department of Microbiology, Immunology, and Cell Biology, West Virginia University, School of Medicine, Morgantown, West Virginia, USA
    Search for more papers by this author
  • Jonathan Benincosa,

    1. Mary Babb Randolph Cancer Center, Robert C. Byrd Health Sciences Center, West Virginia University, School of Medicine, Morgantown, West Virginia, USA
    Search for more papers by this author
  • James Fortney,

    1. Department of Pediatrics, West Virginia University, School of Medicine, Morgantown, West Virginia, USA
    Search for more papers by this author
  • Laura F. Gibson Ph.D.

    Corresponding author
    1. Department of Pediatrics, West Virginia University, School of Medicine, Morgantown, West Virginia, USA
    2. Department of Microbiology, Immunology, and Cell Biology, West Virginia University, School of Medicine, Morgantown, West Virginia, USA
    3. Mary Babb Randolph Cancer Center, Robert C. Byrd Health Sciences Center, West Virginia University, School of Medicine, Morgantown, West Virginia, USA
    • P.O. Box 9214, Department of Pediatrics, School of Medicine, WV University, Morgantown, West Virginia, 26506, USA. Telephone: 304-293-5820; Fax: 304-293-4341
    Search for more papers by this author


Dose-escalated chemotherapy has proven utility in a variety of treatment settings, including preparative regimens before bone marrow or hematopoietic stem cell transplantation. However, the potential damage imposed by aggressive regimens on the marrow microenvironment warrants further investigation. In the present study, we tested the hypothesis that dose-escalated chemotherapy, with etoposide as a model chemotherapeutic agent, activates the transforming growth factor beta-1 (TGF-β1) signaling pathway in bone marrow stromal cells. After high-dose etoposide exposure in vitro, Smad3 protein was phosphorylated in a time-and dose-dependent manner in marrow-derived stromal cells, coincident with the release of active and latent TGF-β1 from the extracellular matrix. Phosphorylation was modulated by p38 kinase, with translocation of Smad3 from the cytoplasm to the nucleus subsequent to its phosphorylation. Etoposide induced activation of TGF-β1 followed the generation of reactive oxygen species and required matrix metalloproteinase-2 (MMP-2) protein availability. Chemotherapy effects were diminished in MMP-2−/− knockout stromal cells and TGF-β1 knockdown small interfering RNA–transfected stromal cells, in which phosphorylation of Smad3 was negligible after etoposide exposure. Stable transfection of a human MMP-2 cDNA into bone marrow stromal cells resulted in elevated phosphorylation of Smad3 during chemotherapy. These data suggest TGF-β1/p38/Smad3 signaling cascades are activated in bone marrow stromal cells after dose-escalated chemotherapy and may contribute to chemotherapy-induced alterations of the marrow microenvironment.


The bone marrow microenvironment serves as the primary site of normal postnatal hematopoiesis and supports hematopoietic recovery after myelosuppressive chemotherapy or irradiation-induced injury of the immune system [1, 2]. Hematopoietic reconstitution requires efficient migration of transplanted stem/progenitor cells to the bone marrow and relocation to stromal cell niches in this microenvironment [3]. The effects of preparative regimens on the marrow microenvironment remain an area requiring further investigation. The assumption that aggressive chemotherapy spares the bone marrow microenvironment grows increasingly more suspect because dose escalation of chemotherapy reveals unexpected problems with hematopoietic recovery [4, 5]. The dilemma remains maintaining efficacy of tumor eradication while reducing damage to the microenvironment.

Of the signaling molecules in the bone marrow microenvironment that may be involved in chemotherapy-induced bone marrow damage, transforming growth factor (TGF)-β1 is specifically noteworthy. TGF-β1 regulates a variety of biological responses, including angiogenesis, chemotaxis, cell-cycle progression, differentiation, and apoptosis of target cells in a context- and cell-specific manner [6, 7]. TGF-β1 is also involved in regulating extracellular matrix (ECM) remodeling, collagen gene expression, and degradation of matrix proteins during the processes of tissue injury and repair [6, 7]. Upregulated expression or activation of TGF-β1 at sites of injury is associated with proliferation of fibroblasts, progressive fibrosis, and subsequent organ dysfunction in diverse systems, including kidney, liver, and lung [811]. In contrast to its promotion of mesenchymal cell proliferation and survival, TGF-β1 is a potent inhibitor of hematopoietic stem cell proliferation [12, 13].

TGF-β1 is initially synthesized as a large precursor that is processed to a mature protein during secretion. After secretion, mature TGF-β1 (25 kD) noncovalently associates with its N-terminal pro-peptide, the 75-kD latency-associated protein (LAP) [14]. The TGF-β1–LAP complex predominantly binds to a latent TGF-β1–binding protein (LTBP), which mediates deposition of the latent complex (230 and 195 kD) to the ECM [14]. Release of mature TGF-β1 from the latent complex can be accomplished by different mechanisms, such as proteolytic cleavage of LAP by plasmin [15], deglycosylation of LAP [16], or interaction with thrombospondin-1 [17], platelet [18], or integrin α4, β6 [19]. After TGF-β1 ligand binding, TGF-β1 receptor II recruits and activates TGF-β1 receptor I, which in turn phosphorylates and activates the R-Smads, including Smad2 or Smad3 [20]. Phosphorylated R-Smads homodimerize, form a transcriptional complex with Smad4, and translocate into the nucleus to regulate target gene expression [20].

Data presented in the current study suggest that the TGF-β1/p38/Smad3 signaling cascades are activated through reactive oxygen species (ROS)–mediated matrix metalloproteinase-2 (MMP-2) activity in bone marrow stromal cells during etoposide chemotherapy. Increased availability of active TGF-β1 has the potential to alter stromal cell function through regulation of diverse Smad-driven gene expression in stromal cells. Moreover, release of TGF-β1 from ECM during chemotherapy may also directly regulate growth and proliferation of transplanted hematopoietic stem cells. This in vitro model provides a setting in which we can further delineate the effects of chemotherapy on marrow stromal cells and evaluate the role of TGF-β1 in influencing hematopoietic recovery after transplantation.

Materials and Methods

Cell Cultures

HS-27A human bone marrow–derived stromal cells [21] (American Type Culture Collection [ATCC, Manassas, VA,] no. CRL-2496) were maintained in alpha-modification of Eagle's medium (α-MEM) (GIBCO, Grand Island, NY, with supplements as recommended by the ATCC. Ped604, P148, and P156 are primary bone marrow stromal cells derived from consenting donors with the approval of the West Virginia University Institutional Review Board. Establishment of bone marrow stromal cells and their characterization by our laboratory have been previously described in detail [22]. Bone marrow aspirates from MMP-2 knockout (MMP-2−/−, KO) or wild-type (MMP-2+/+, WT) C57BL/6 mice were generously provided by Dr. Farrah Kheradmand [23], Baylor College of Medicine. Murine bone marrow stromal cell line S-10 (provided by Dr. Kenneth Dorshkind, University of California at Los Angeles, CA) and stromal cell–dependent and interleukin (IL)-7–dependent murine pro-B cell line C1.92 (provided by Dr. Kenneth Landreth, West Virginia University, Morgantown, WV) have been previously described [24].

Chemotherapeutic and Other Chemical Agents

Etoposide (VP-16; Bristol Laboratories, Princeton, NJ, was stored at a stock concentration of 33.98 mM. A final concentration of 100 μM was used to approximate pre-transplant clinical treatment [25]. Cytarabine (Ara-C; Sigma, St. Louis, was reconstituted at 10 mg/ml and stored at −20°C. Doxorubicin (3 mM) was purchased from Gensia Sicor Pharmaceuticals (Irvine, CA,, and 4-hydroperoxycyclophosphamide (4-HC) (10 mg/ml) was a gift from Dr. T. Ball (University of California, San Diego). Danunorubicin (Sigma), Melphalan (Sigma), and Vincristine (Sigma) were reconstituted at 10 μg/μl immediately before use. Experimental concentrations of chemotherapeutic drugs are noted in the appropriate figure legends.

The MMP-2 inhibitor cis-9-octadecenoyl-N-hydroxylamide (OA-Hy), Erk1/2 kinase inhibitor U0126, p38 kinase inhibitor SB220025, and JNK/SAPK inhibitor SP600125 were purchased from Calbiochem (La Jolla, CA, ROS scavenger N-acetyl-cysteine (NAC) was purchased from Sigma. Recombinant active MMP-2 and pro–MMP-2 were purchased from BioMol (Plymouth Meeting, PA, and Calbiochem, respectively. Recombinant human TGF-β1 (rh-TGF-β1) was obtained from R&D Systems (Minneapolis,

In the indicated experiments, stromal cells were preincubated with 1 μM OA-Hy for 30 minutes or 250 ng/ml active or pro–MMP-2 for 15 minutes before exposure to chemotherapy for 1 hour. For in vitro activation of MMP-2, pro–MMP-2 was incubated with 10 μM hydrogen peroxide (H2O2) (8.8 N; Sigma) at 37°C for 15 minutes immediately before use. When indicated, stromal cells were pretreated with 10 μM U0126, 20 μM SB220025, or 5 μM SP600125 for 30 minutes before etoposide exposure for an additional 1 hour.

Transfection of Murine Stromal Cells with Human MMP-2

A 2,119-bp EcoRI cDNA fragment encoding the full-length human MMP-2 was cut from the entry plasmid pBR322-MMP-2-amp(+) (ATCC no. 65016) and inserted into the multiple cloning site of the mammalian expression plasmid pUSE-CMV-neo (Upstate, Lake Placid, NY, Subcloning was carried out after purification using the MiniElute gel purification kit (Qiagen Sciences, Germantown, MD, with the cDNA ligated with T4 DNA ligase (Invitrogen, Carlsbad, CA, Transfection of stromal cells with the pUSE-MMP-2-neo construct or its empty vector control pUSE-CMV-neo was conducted following the protocols described previously [26, 27]. Briefly, 16 hours before transfection, S-10 stromal cells were cultured in α-MEM supplemented with 5% fetal bovine serum with no antibiotics (transfection growth medium [TGM]). Plasmid DNAs and Lipofectamine 2000 (Invitrogen) were diluted with Opti-MEM (Invitrogen) and mixed at variable ratios at room temperature for 20 minutes. Murine stromal S-10 cells were transfected with either the vector or MMP-2 construct followed by G418 selection (0.5 mg/ml). Stable clones expressing both the human MMP-2 and neomycin-resistant gene product neomycin phosphotransferase II (NPT II) or the NPT II alone were selected for further experiments. These are designated SM-8 and SV-2, respectively.

TGF-β1 Knockdown by siRNA

For transient TGF-β1 small interfering RNA (siRNA) transfection, Lipofectamine 2000 and nontargeting double-stranded RNA (dsRNA) control or TGF-β1 knockdown siRNA (Dharmacon, Boulder, CO, were diluted and combined. HS-27A and P148 human stromal cells were cultured in TGM overnight and transfected with 50–150 nM TGF-β1 knockdown siRNA or control dsRNA. Control siRNA was consistently used at the highest concentration of TGF-β1 siRNA in all experiments. Forty-eight hours after transfection, TGM was replaced with serum-free medium and stromal cells were treated with 100 μM etoposide for 1 hour. Stromal cell supernatants and cell pellets were collected for enzyme-linked immunosorbent assay (ELISA), zymography, or Western blot analyses.

Quantitation of Active and Total TGF-β1 by ELISA

Quantitation of the release of active and total TGF-β1 from stromal cell ECM during chemotherapy was measured by ELISA according to the recommendations of the manufacturer (R&D Systems). Briefly, confluent stromal cells were plated in six-well plates in serum-free medium overnight. After exposure to 0–100 μM etoposide for 1 hour or 100 μM etoposide for 5 minutes to 6 hours, stromal cell supernatants were collected. Ten minutes before each treatment, 0.5 μg/ml anti–TGF-β1 antibody was added to each well to stabilize the released TGF-β1. Supernatants were acidified with 1.0 N HCl solution and neutralized with 1.2 N NaOH/0.5 M HEPES solution immediately before assay to measure total TGF-β1. For quantitation of active/free TGF-β1, supernatants were directly subjected to ELISA without acid activation. Both acidified and nonacidified samples were measured in triplicate, and colorimetric development was determined at 450 nm with the correction wavelength at 540 nm on a multiwell plate reader (Bio-Tek Instruments, Inc., Winooski, VT,

Antibodies and Western Blot Analysis

Rabbit polyclonal anti-phospho-Smad3 (Ser433/435), rabbit monoclonal anti-phospho-p38 kinase (Thr180/Tyr182), rabbit monoclonal anti-phospho-Erk1/2 mitogen-activated protein kinase (MAPK) (Thr202/Tyr204), and rabbit monoclonal anti-phospho-JNK/SAPK (Thr183/Tyr185) were purchased from Cell Signaling Technology (Beverly, MA, Mouse monoclonal anti-p38 kinase and rabbit polyclonal anti-JNK2 antibodies were also from Cell Signaling Technology. Rabbit polyclonal anti-Erk2 and rabbit polyclonal anti-Erk1 were from Santa Cruz Biotechnology (Santa Cruz, CA, Mouse monoclonal anti-Smad3 was from BD Biosciences Transduction Laboratories (San Diego, Mouse monoclonal anti-human TGF-β1, LAP/TGF-β1, and LTBP-1/TGF-β1 antibodies were obtained from R&D Systems. Rabbit polyclonal anti-NPT II antibody was purchased from Upstate.

Cells were lysed in complete cell lysis buffer (CCLB) (50 mM Tris-HCl, pH7.4, 150 mM NaCl, 1% Triton X-100, 0.25% Na-deoxycholate, 1 mM EDTA, and 1 mM NaF, 1 mM dithiothreitol (DTT), 1 mM phenylmethylsulphonylfluoride, 1 mM activated Na3VO4, 1 μg/ml aprotinin, 1 μg/ml leupeptin, and 1 μg/ml pepstatin) on ice for 15 minutes. After centrifugation at 14,000 rpm for 15 minutes, supernatants were collected and protein concentration was determined using the bicinchoninic acid (BCA) protein assay (Pierce, Rockford, IL, Proteins were resolved on SDS-PAGE gels and transferred to nitrocellulose membranes. Membranes were blocked in TBS 5%/nonfat dry milk 0.05% Tween-20 and probed with the indicated primary antibodies. After incubation with horse radish peroxidase–conjugated secondary antibodies, signal was visualized using enhanced chemiluminescence reagents (Amersham, Piscataway, NJ,

Immunoprecipitation of TGF-β1

Confluent stromal cells were rinsed with serum-free α-MEM and then recultured in serum-free media. Anti-human TGF-β1, LAP/TGF-β1, and LTBP-1/TGF-β1 antibodies were added at a final concentration of 3 μg/ml for 15 minutes before addition of etoposide for 1 hour. Supernatants were collected and combined with protein A/G agarose beads (Santa Cruz Biotechnology) at 4°C for 4 hours. The immunoprecipitates were washed with CCLB and heated to 100°C for 5 minutes before separation on SDS-PAGE gels under both reducing and nonreducing conditions.

Heavy and light chains served as the loading controls for the immunoprecipitation (IP) experiments.

Gelatin Zymography

Bone marrow stromal cell supernatants were collected after 100 μM etoposide treatment in serum-free α-MEM. Supernatants were concentrated 10-fold using Amicon Ultra-15 centrifugal filters (Millipore, Billerica, MA, spun at 3000g for 95, 30, and 10 minutes, respectively, at room temperature. For gelatinolytic analysis of cell lysates, cell pellets were lysed in CCLB without NaVO3, NaF, EDTA, and DTT. After quantitation of supernatant and cell lysate protein by the BCA protein assay, samples were resolved in 10% SDS-PAGE gels containing 1% gelatin (Sigma) under nonreducing conditions. After electrophoresis, gels were incubated for 30 minutes in 2.5% Triton X-100 (Mallinckrodt, Inc., Paris, KY, and subsequently incubated overnight at 37°C in 1 × developing buffer (1.2% Tris Base, 6.3% Tris HCl, 11.7% NaCl, 0.7% CaCl, 0.2% Brij 35). Gels were then stained with 0.5% Coomassie Blue R-250 (Bio-Rad Laboratories, Richmond, CA, for 30 minutes at room temperature and then destained (50% methanol, 10% acetic acid, 40% dH20) until clear bands were detected, indicative of active MMP-2.

Detection of Intracellular ROS by Flow Cytometry

Detection of intracellular ROS generation by flow cytometry was performed as previously described [28]. Briefly, confluent stromal cells were pretreated with 10 μM carboxyl-H2DCF-DA (Molecular Probes, Eugene, OR, for 30 minutes followed by etoposide exposure for various times. Nonfluorescent carboxyl-H2DCF-DA was hydrolyzed to H2DCF, which is oxidized in the presence of H2O2 and emits fluorescence detected in the FL1-H channel. Cells were trypsinized, rinsed with phosphate-buffered saline (PBS), and immediately run on a BD Biosciences FACScan. Data were analyzed and processed with CellQuest Pro software (Becton, Dickinson and Company, Franklin Lakes, NJ,

Confocal Microscopy

Stromal cells were cultured on coverslips and exposed to etoposide for 30 minutes to 6 hours. After fixation with methanol/acetone (1:1) for 30 minutes at room temperature, stromal cells were incubated with 3 μg/100 μl antiphospho-Smad3 antibody or isotype control antibody for 3 hours. After three washes with autoclaved PBS, cells were incubated with goat anti-rabbit IgG-fluorescein isothiocyanate (FITC) (Southern Biotechnology Associates, Birmingham, AL, for 1 hour. Propidium iodide (PI) (5 μg/100 μl) was used to counterstain the nuclei. Coverslips were mounted onto slides with Fluormount-B (Fisher Scientific, Orangeburg, NY, and evaluated by confocal microscopy (LSM510; Carl Zeiss, Jena, Germany,

Pro-B Cell Adhesion Assays

Stromal cells were plated in 96-well plates and exposed to 0–5 ng/ml rh-TGF-β1 for 72 hours. Cells were thoroughly rinsed with fresh medium three times before establishment of C1.92 pro-B/stromal cell coculture. Before coculture, C1.92 pro-B cells were labeled with the fluorescent dye PKH-26 (Sigma) for 3 minutes and then washed with medium. Five × 105 C1.92 cells were cocultured with stromal cells for an additional 2 hours. Nonadherent C1.92 cells were removed by three PBS rinses. Ninety-six–well plates were then analyzed on a multiwell fluorimetric reader (CytoFluor, Perseptive Biosystems, Foster City, CA, to quantitate fluorescence as a measure of stromal cell–bound C1.92 cells. Stromal cells alone were included to quantitate any background fluorescence.

Pro-B Cell Proliferation Assay

The effect of TGF-β1 on the ability of murine S10 or human-derived stromal cells to support pro-B cell proliferation was investigated by exposing 100% confluent stromal cell layers to increasing doses of rh-TGF-β1 (0–5 ng/ml) for 72 hours in 96-well plates. After exposure of stroma to TGF-β1 in vitro, stromal cells were thoroughly rinsed from culture, and 5 × 105 pro-B cells per ml were added to each well in fresh α-MEM. The proliferative response of pro-B cell clone C1.92 to murine stromal cell line S10 has been well characterized [24]; therefore, this combination of cells is particularly informative in determining the effect of TGF-β1 on the ability of stromal cells to support pro-B cell expansion. Recombinant murine IL-7 (Biosource, Camarillo, CA, (25 U/ml) was included in all samples. Wells were pulsed with 1 μCi 3H-TdR 16 hours after the addition of C1.92 and harvested onto glass wool fiber strips 6 hours later. Incorporated radioactivity was determined by liquid scintillation counting (LKB-Wallac Model 1410, Gaithersburg, MD, in an aqueous fluor (Biosafe-II; Research Products International, Mount Prospect, IL, Control wells of untreated stroma were included in each experiment.

Statistical Analysis

Data presented were expressed as mean ± SEM for triplicate samples. Statistic significance was determined using the Student's t-test. p values less than .05 were considered significant.


Chemotherapy Activates Smad3 Through Phosphorylation at Serines 433/435 in Human Bone Marrow–Derived Stromal Cells

To investigate the phosphorylation of Smad3 in stromal cells after chemotherapeutic stimulation, stromal cells were treated either for different times or with various concentrations of etoposide. Exposure of stromal cells to etoposide resulted in phosphorylation of Smad3 at serines 433/435 in a time-dependent (Fig. 1A) and dose-dependent (Fig. 1B) manner. Etoposide induced a rapid elevation of phospho-Smad3 signal as early as 30 minutes, which was sustained for approximately 6–7 hours (Fig. 1A). After the transient increase, phospho-Smad3 levels diminished for up to 24 hours in the presence of chemotherapy. Etoposide, melphalan, vincristine, daunorubicin, doxorubicin, 4-hydroperocyclophosphomide, and Ara-C induced Smad3 phosphorylation in stromal cells to varying degrees (Fig. 1C). Treatment of stromal cells with recombinant TGF-β1 served as a positive control and induced the most pronounced phosphorylation of Smad3. Total Smad3 protein remained unchanged and served as the lane loading control.

Figure Figure 1..

Chemotherapy activates Smad3 through phosphorylation at serines 433/435 in human bone marrow–derived stromal cells. Western blot analyses of HS-27A stromal cells treated with (A) 100 μM etoposide for the indicated time points, (B) various concentrations of etoposide for 1 hour, or (C) 3 ng/ml transforming growth factor -β1 (TGF-β1), 100 μM etoposide (VP-16), 200 μg/ml melphalan (Mel), 20 μg/ml vincristine (VCR), 100 μg/ml daunorubicin (DNR), 100 μM doxorubicin (DOX), 100 μg/ml 4-hydroperoxylcyclophosphamide (4-HC), or 100 μg/ml Ara-C for 1 hour. Membranes were probed with anti-phospho-Smad3 (P-Smad3, ser433/435) and then stripped and reprobed with total Smad3 (T-Smad3)–specific antibodies.

Chemotherapy-Induced Smad3 Phosphorylation Is Mediated by TGF-β1

To investigate the potential involvement of TGF-β1 in phosphorylation of Smad3 in bone marrow stromal cells during chemotherapy, HS-27A stromal cells were exposed to etoposide for different times or at various concentrations. Quantitative analysis of TGF-β1 by ELISA was performed using the cell supernatants after treatment. To better distinguish the free (active) and latent (total) TGF-β1 released, nonacidification and acidification of the stromal supernatants were simultaneously used before assay as described. Exposure of stromal cells to chemotherapy resulted in elevated TGF-β1 release from stromal cells both in a time-and dose-dependent manner (Fig. 2A). Chemotherapy rapidly induced the release of active and latent forms of TGF-β1 from stromal cell ECM. In our stromal cell model, active TGF-β1 constituted approximately 5%–9% of the total TGF-β1 pool during each treatment phase. Activation of TGF-β1 preceded phosphorylation of Smad3 with initial increases rapidly after etoposide treatment of 15 minutes and further elevation at 1 hour. Activation of TGF-β1 in stromal cells was transient, as longer than 1 hour exposure of stromal cells to chemotherapy correlated with gradual regression of active and total TGF-β1 to the baseline level.

Figure Figure 2..

Chemotherapy-induced Smad3 phosphorylation is mediated by TGF-β1. (A): Quantitative analysis of the release of active and total TGF-β1 from HS-27A stromal cell supernatants exposed to eto-poside at 0–100 μM for 1 hour (left graph) or at 100 μM for 0–6 hours (right graph). Bars marked with * or # indicate significant differences compared with untreated controls (p < .05). (B): Immunoprecipitation of human TGF-β1 from HS-27A and P148 supernatants after exposure of stromal cells to the indicated concentrations of etoposide for 1 hour. Supernatants were immunoprecipitated with anti–TGF-β1 and run under reducing conditions. Western blots were probed with anti–TGF-β1 antibody. Recombinant human TGF-β1 served as the molecular-size control. (C): Immunoprecipitation of TGF-β1 from HS-27A cell supernatants preincubated with 3 μg of anti–TGF-β1, anti–LAP/TGF-β1, or anti–LTBP-1/TGF-β1 antibodies followed by exposure to 100 μM etoposide for 1 hour. Samples were run under both reducing and nonreducing conditions, and Western blots were probed with anti-human TGF-β1 antibody. Abbreviations: Ig, immunoglobulin; LAP, latency-associated protein; LTBP, latent TGF-β1–binding protein; TGF-β1, transforming growth factor beta-1.

IP of TGF-β1 from the stromal cell supernatants indicated that baseline TGF-β1 in untreated stromal cell supernatants was negligible, with increased TGF-β1 immunoprecipitated from etoposide-treated stromal cell supernatants in a dose-dependent fashion (Fig. 2B).

Because the anti–TGF-β1 antibody we used for IP of TGF-β1 may recognize both active and latent forms of TGF-β1, we performed additional IP experiments with antibodies recognizing the free and total TGF-β1 (i.e., anti–TGF-β1), the small latent complex (i.e., anti–LAP/TGF-β1), or the large latent complex (anti–LTBP-1/TGF-β1) to further address this issue. As shown in Figure 2C, under nonreducing electrophoretic conditions, the major forms of TGF-β1 activated via etopo-side treatment are the 230- and 195-kD large latency complexes (i.e., LTBP-1/LAP/TGF-β1) as immunoprecipitated by anti–TGF-β1, anti–LAP/TGF-β1, and anti–LTBP-1/TGF-β1 antibodies. In addition, a 100-kD band, which represents the small latency complex (i.e., LAP/TGF-β1), and a 75-kD LAP band were also detected in etoposide-treated samples. When the same samples were electrophoresed under reducing conditions, the high-molecular-weight large and small latency complexes were partially dissociated, and two bands of molecular size of 195 kD (LTBP-1/LAP- TGF-β1) and 25 kD (TGF-β1) were observed.

Disruption of the Availability of TGF-β1 Blocks the Signal Transduction Initiated by Chemotherapy

To better understand the role of TGF-β1 in mediating chemotherapy-triggered signals during bone marrow damage, marrow-derived stromal cells were treated with chemotherapeutic agents in the presence or absence of anti–TGF-β1 neutralizing antibody. Etoposide, melphalan, and 4-HC promoted phosphorylation of Smad3 when cells were pretreated with the isotype control antibody, whereas phosphorylation of Smad3 was diminished in the presence of TGF-β1 neutralizing antibody (Fig. 3A).

Figure Figure 3..

Disruption of the availability of transforming growth factor-β1 (TGF-β1) blocks the signal transduction initiated by chemotherapy. (A): Western blot analysis of HS-27A stromal cells treated with 100 μM etoposide, 200 μg/ml melphalan, or 100 μg/ml 4-hydroperoxylcyclophosphamide (4-HC) for 1 hour. (B): Enzyme-linked immunosorbent assay of released total TGF-β1 from HS-27A and P148 stromal cells transfected with the indicated concentrations of TGF-β1 knockdown small interfering RNA (siRNA) or control double-stranded RNA (dsRNA) for 48 hours followed by exposure to 100 μM etoposide for 1 hour (upper panel). Bars marked with * or # indicate significant differences compared with control dsRNA transfections (p < .05). Western blot analysis of Smad-3 phosphorylation using the same TGF-β1 siRNA transfection cell lysates is shown in the lower panel.

This prompted us to more specifically test whether chemotherapy-induced effects on marrow stromal cells could be disrupted through downregulation of TGF-β1 expression. Human HS-27A and P148 stromal cells were transiently transfected with TGF-β1 knockdown siRNA before exposure to etoposide. TGF-β1 targeting siRNA transfection diminished the amount of total TGF-β1 release induced by etoposide treatment in a concentration-dependent manner compared with control dsRNA transfection (Fig. 3B, upper panel). Complete loss of TGF-β1 release occurred when stromal cells were exposed to 150 nM siRNA in the presence of chemotherapy. Consistent with the diminished availability of TGF-β1 presented in the supernatants, phosphorylation of Smad3 was also reduced after transfection of stromal cells with various concentrations of TGF-β1 siRNA during chemotherapy (Fig. 3B, lower panel).

Chemotherapy-Induced MMP-2 Activity Is Required for Activation of Latent TGF-β1

Gelatin zymography revealed that MMP-2 activity was elevated in S10 stromal cell supernatants after etoposide exposure as early as 5 minutes and increased further at 30 to 60 minutes (Fig. 4A). Inhibition of MMP-2 activity by OA-Hy diminished phosphorylation of Smad3 after etoposide treatment of stromal cells (Fig. 4B). To determine whether MMP-2 was required for bone marrow stromal cell activation of TGF-β1, we established stromal cells from MMP-2−/− knockout mice. Etoposide, Melphalan, or 4-HC exposure induced Smad3 phosphorylation in murine MMP-2+/+ stromal cells, whereas phospho-Smad3 signals were less pronounced in MMP-2−/− stromal cells (Fig. 4C). Addition of active MMP-2 partially restored treatment-induced phospho-Smad3 signals in MMP-2−/− cells and further increased phosphorylation of Smad3 in MMP-2+/+ stromal cells treated with etoposide (Fig. 4D).

Figure Figure 4..

Chemotherapy-induced MMP-2 activity is required for activation of latent TGF-1. (A): Gelatin zymography analysis of supernatants from S-10 stromal cells treated with 100 μM etoposide for 0 to 6 hours. (B): Western blot analysis of HS-27A and Ped604 stromal cells treated with 1 μM MMP-2 inhibitor OA-Hy for 30 minutes before exposure to 100 μM etoposide for 1 hour. (C): Western blot analysis of murine C57BL/6 MMP-2 knockout stromal cells (MMP-2−/−, KO) or wild-type stromal cells (MMP-2+/+, WT) treated with 100 μM etoposide, 200 μg/ml melphalan, or 100 μg/ml 4-HC for 1 hour. (D): Western blot analysis of MMP-2+/+ and MMP-2−/− stromal cells treated with 100 μM etoposide in the presence or absence of 250 ng/ml recombinant human active MMP-2 for 1 hour. Untreated controls were not exposed to either recombinant MMP-2 or etoposide. (E): Gelatinolytic analysis of supernatants and cell lysates from S-10 parental, vector, or MMP-2 transfected S-10 stromal cells treated with 100 μM etoposide for 1 hour and Western blot analysis of Smad3 phosphorylation and NPT II expression using the same transfection samples. Abbreviations: 4-HC, 4-hydroperoxycyclophosphamide; KO, knockout; MMP-2, matrix metalloproteinase-2; NPT II, neomycin phosphotransferase II; OA-Hy, cis-9-octadecenoyl-N-hydroxylamide; TGF, transforming growth factor; WT, wild-type.

To further investigate the role of MMP-2 in mediating activation of TGF-β1 in marrow stromal cells during chemotherapy, S-10 murine stromal cells transfected with a human MMP-2 construct or vector control were established. Stromal cell clones with comparable expression of the neomycin resistance gene NPT II were selected for further experiments. As shown in Figure 4E, although no substantial differences were observed between S-10 parental and SV-2 vector–transfected cells in terms of activation of Smad3 and MMP-2 during treatment, overexpression of MMP-2 in SM-8 stromal cells increased baseline and etoposide-induced Smad3 phosphorylation.

Activation of Latent MMP-2 by Chemotherapy Requires the Generation of Reactive Oxygen Species

Because MMP-2 exists largely as a latent form in stromal cell matrix, we next sought to explore the mechanism underlying the activation of pro–MMP-2 during etoposide chemotherapy. ROS was generated after etoposide treatment and was required for conversion of pro–MMP-2 to its active form. Etoposide rapidly induced production of intracellular ROS in HS-27A stromal cells as early as 5 minutes after etoposide exposure, which preceded activation of MMP-2 and TGF-β1 (Fig. 5A, upper panel).

Figure Figure 5..

Activation of latent MMP-2 by chemotherapy requires the generation of ROS. (A): Flow cytometric analysis of intracellular ROS in HS-27A stromal cells treated with 100 μM etoposide for 0–4 hours (upper panel) or HS-27A stromal cells treated with the indicated chemotherapeutic agents identical to those shown in Figure 1C for 1 hour (lower panel). Untreated control stromal cells in the lower panel are indicated by the solid histogram. Ara-C–treated stromal cell ROS overlays the untreated control histogram. (B): Western blot analysis of HS-27A or P148 stromal cells pretreated with 20 mM NAC overnight before exposure to 100 μM etoposide (VP-16) for 1 hour. (C): Western blot analysis of HS-27A stromal cells treated with recombinant pro–MMP-2 that was activated in vitro. Cells were exposed to 0–500 ng/ml activated MMP-2 for 1 hour. (D): Western blot analysis of MMP-2+/+ and MMP-2−/− stromal cells treated with 10 μM H2O2 in the presence or absence of 250 ng/ml pro–MMP-2 for 1 hour. Abbreviations: 4-HC, 4-hydroperoxylcyclophosphamide; DNR, daunorubicin; DOX, doxorubicin; KO, knockout; Mel, melphalan; MMP-2, matrix metalloproteinase-2; NAC, N-acetyl-cysteine; ROS, reactive oxygen species; VCR, vincristine; WT, wild-type.

Comparable to the activation of MMP-2 and TGF-β1, ROS generation is also a transient event during chemotherapy in our stromal cell model. The mean fluorescence intensity emitted by oxidized 2′,7-dichlorodihydrofluorescein diacetate in etoposide-treated stromal cells was increased greater than twofold to threefold in all lines evaluated compared with untreated controls. Uniquely, Ara-C did not stimulate stromal cell production of H2O2 during short-term (1-hour) chemotherapy (Fig. 5A, lower panel). Reduction of intracellular ROS accumulation with the hydroxyl radical scavenger NAC reduced phospho-Smad3 in stromal cells treated with etoposide (Fig. 5B).

To confirm the role of ROS in the activation of MMP-2, pro–MMP-2 was activated in vitro by hydrogen peroxide. As shown in Figure 5C, treatment of HS-27A stromal cells with in vitro–activated MMP-2 induced phosphorylation of Smad3 in a dose-dependent manner. Because inhibition of extracellular MMP-2 activity and reduction of intracellular ROS both disrupted etoposide-induced Smad3 phosphorylation, we sought to determine which one was the initiating factor in modulating TGF-β1/Smad3 signaling. Hydrogen peroxide–induced phosphorylation of Smad3 only occurred in MMP-2+/+ but not MMP-2−/− cells, whereas in the presence of pro–MMP-2, oxidative stress led to phosphorylation of Smad3 in MMP-2−/− cells (Fig. 5D).

P38 Mediates Etoposide-Induced Smad3 Phosphorylation in Bone Marrow Stromal Cells

Smad3 was not directly phosphorylated by TGF-β1 receptor I in a classic fashion but appeared to be regulated by p38 MAPK in this specific setting. As indicated in Figure 6A, all the chemotherapeutic drugs evaluated in this study activated p38 kinase. Chemotherapy induced phosphorylation and activation of both Erk1/2 and p38 kinases in HS-27A cells (Fig. 6B); however, inhibition of Erk1/2 MAPK with U0126 did not result in diminished phosphorylation of Smad3. In contrast, interruption of p38 kinase activity with SB220025 blocked etoposide-triggered Smad3 phosphorylation. JNK/SAPK was not involved in chemotherapy-induced activation of TGF-β1 signaling in bone marrow stromal cells.

Figure Figure 6..

P38, but not Erk1/2 or JNK kinase, is involved in mediating etoposide (VP-16)–induced Smad3 phosphorylation. (A): Western blot analysis of HS-27A stromal cells treated with TGF-β1 or the same chemotherapeutic agents shown in Figures 1C and 5A (lower panel). (B): Western blot analysis of HS-27 stromal cells pretreated with vehicle, 10 μM Erk1/2 kinase inhibitor U0126, 20 μM p38 kinase inhibitor SB220025, or 5 μM JNK/SAPK inhibitor SP600125 followed by etoposide exposure for 1 hour. Blot was stripped and reprobed with antibodies specific for the proteins indicated. Abbreviations: 4-HC, 4-hydroperoxylcyclophosphamide; DMSO, dimethyl sulfoxide; DNR, daunorubicin; DOX, doxorubicin; Mel, melphalan; TGF-β1, transforming growth factor-beta 1; VCR, vincristine.

Etoposide Treatment Results in Redistribution of Phosphorylated Smad3 Protein in Human Stromal Cells

Changes in cellular distribution of Smad3 protein after etoposide-induced phosphorylation were evaluated (Fig. 7). The phospho-Smad3 signal was negligible in untreated stromal cells, with only the PI-counterstained cell nuclei clearly detected. Cytoplasmic Smad3 was rapidly phosphorylated in response to etoposide stimulation as early as 30 minutes. Longer exposure of stromal cells to etoposide induced a gradual redistribution and accumulation of Smad3 protein in nucleus. After approximately 4 hours of etoposide treatment, most phospho-Smad3 had translocated into the stromal cell nuclei.

Figure Figure 7..

Etoposide treatment results in phosphorylation and subsequent nuclear localization of Smad3 protein in human stromal cells. Double-channel confocal microscopy analysis of HS-27A stromal cells treated with 100 μM etoposide for up to 6 hours. Cells were double-stained with 5 μg/100 μl propidium iodide (PI) (rhodamine, red signal) and 3 μg/100 μl of antiphospho-Smad3 antibody followed by staining with fluorescein isothiocyanate (FITC)–conjugated anti-rabbit IgG secondary antibody (green signal). Panels on the right represent merged images of cytosolic and nuclear phospho-smad3 staining. Original magnifications ×200.

Recombinant TGF-β1 Activates Smad3 and Impairs Stromal Cell Support of Pro-B Cell Adhesion and Proliferation

To characterize the response of stromal cells to TGF-β1 exposure, human primary P156 stromal cells were treated with rh-TGF-β1 (Fig. 8A). TGF-β1 rapidly induced phosphorylation of Smad3 in P156 stromal cells in a time-dependent manner with elevated Smad3 phosphorylation as early as 30 minutes and decreased phospho-Smad3 signals thereafter (Fig. 8A, upper panel). In contrast to the Smad3 activation pattern induced by chemotherapy, rh-TGF-β1 treatment resulted in the most pronounced Smad3 phosphorylation at 1 and 5 ng/ml of TGF-β1; however, 10 and 20 ng/ml resulted in diminished phospho-Smad3 signals (Fig. 8A, lower panel).

Figure Figure 8..

Recombinant transforming growth factor (TGF)-β1 activates Smad3 and impairs stromal cells support of pro-B cell adhesion and proliferation. (A): Western blot analysis of Smad3 phosphorylation of HS-27A stromal cells treated with 3 ng/ml rh-TGF-β1 for the indication time points or the indicated concentrations of rh-TGF-β1 for 24 hours. (B): Adhesion of C1.92 pro-B cells labeled with PKH-26 and cocultured on P156 and S-10 stromal cells pretreated with the indicated concentrations of rh-TGF-β1 for 72 hours. Bars marked with * or # indicate significant differences compared with untreated controls (p < .05). (C):3H-thymidine incorporation of C1.92 pro-B cells cocultured on P156 and S-10 stromal cells pretreated with the indicated concentrations of rh-TGF-β1 for 72 hours. Bars marked with * or # indicate significant differences compared with untreated controls (p < .05).

To explore the functional consequences of activation of TGF-β1/Smad3 signaling during bone marrow damage, bone marrow stromal cells were treated with rh-TGF-β1 followed by coculture with C1.92 hematopoietic stem cells. Stromal cells pretreated with TGF-β1 diminished the ability to support C1.92 cell adhesion to the stromal cell layer (Fig. 8B). In addition to the diminished adhesion of the pro-B cells, C1.92 cells cocultured on TGF-β1–pretreated human or murine stromal cells had lower cell proliferation (Fig. 8C) compared with those on control stromal cells.


Bone marrow transplantation/hematopoietic stem cell transplantation (HSCT) has been proven to be an effective treatment of many malignancies that are refractory to less aggressive approaches [2932]. During this process, one challenge is maintaining the hematopoietic support capacity of the bone marrow microenvironment. Complications associated with HSCT include aplastic anemia or pancytopenia, delay of hematopoietic recovery, severe immunosuppression-related infections [3335], and secondary myelofibrosis [36, 37]. These observations emphasize the challenge of using high-dose chemotherapy while attempting to maintain function of bone marrow stromal cell niches that support hematopoietic recovery.

We have previously reported that high-dose etoposide exposure, although not reducing the viability of bone marrow stromal cells, resulted in a plethora of functional alterations. Etoposide-treated stromal cells have diminished surface vascular cellular adhesion molecule (VCAM)-1 protein [22] and reduced ability to support chemotaxis of CXCR4-positive progenitor cells [38]. In addition, stromal cell–dependent and IL-7–dependent pro-B cells grown on etoposide-pretreated stromal cells accumulate in G0/G1 phase of the cell cycle and subsequently initiate apoptosis [22]. These observations suggest a variety of treatment-induced stromal cell alterations that potentially influence hematopoietic support capacity.

Because TGF-β1 has a variety of direct inhibitory effects on hematopoietic cells, we hypothesized that etoposide-induced disruption of bone marrow stromal cell support of pro-B cells may result, in part, from activation of TGF-β1 (Fig. 9). Initial studies indicated that bone marrow stromal cells responded to chemotherapeutic exposure with downstream phosphorylation of Smad3. Etoposide treatment rapidly resulted in a dose- and time-dependent phosphorylation of Smad3 protein in bone marrow stromal cells (Figs. 1A, 1B). The response of stromal cells to a variety of drugs, including our model drug etoposide, was similar to that induced by rh-TGF-β1 treatment alone (Fig. 1C). These data suggested an intracellular signal transducer of TGF-β1 was activated in response to chemotherapy and provided indirect evidence that TGF-β1 may be involved in chemotherapy-induced stromal cell alterations. It should be noted that although higher doses of etoposide induced stronger phospho-Smad3 signals, longer exposure of stromal cells to 100 μM etoposide for up to 24 hours did not lead to sustained Smad3 phosphorylation.

Figure Figure 9..

Proposed model for activation of the transforming growth factor (TGF)-β1/p38/Smad3 signaling cascade in bone marrow stromal cells during chemotherapy. After chemotherapy, stromal cell mitochondria generate intracellular reactive oxygen species (ROS), which translocates into the extracellular matrix and oxidizes pro–matrix metalloproteinase-2 (MMP-2) complexes. Active MMP-2 subsequently cleaves and releases TGF-β1 from latent TGF-β1–binding protein, allowing TGF-β1 to bind to its receptor, and initiates phosphorylation of stromal cell p38 kinase. P38 mediates phosphorylation of Smad3 protein, which can subsequently dimerize with Smad4 and translocates into the nuclei to regulate a diverse set of target genes that may influence stromal cell support of hematopoietic cell development.

Several mechanisms may underlie the transient phosphorylation of Smad3. It is generally recognized that the protein phosphatases, specifically protein phosphatase 2A (PP2A), are activated after stress in several cell models [39]. Elevated PP2A activity may subsequently result in dephosphorylation of signaling molecules, such as Smad3, in stromal cells treated with etoposide. However, sustained phosphorylation of Smad3 may not be required to elicit a significant effect. Transit of Smad3 to the nucleus after phosphorylation provides the potential for diverse changes in expression of Smad3-responsive genes and subsequent alteration of stromal cell function.

Because R-Smads can also be phosphorylated/activated in response to other members of the TGF-β1 superfamily [7], we performed several experiments using chemical and genetic approaches to verify that TGF-β1 is specifically involved in chemotherapy-induced Smad3 phosphorylation (Figs. 2, 3). Treatment of stromal cells with chemotherapy resulted in the release of active and total TGF-β1 in the supernatants. The free/active form TGF-β1 only accounts for approximately 5%–9% of the total TGF-β1 pool released from chemotherapy-treated stromal cells. However, MMP-2 may cleave and release LAP from ECM and subsequently release TGF-β1 from LAP or LTBP-1 complexes [40, 41]. Therefore, it can be postulated that TGF-β1 activated through chemotherapy may exert its biological functions in an extended-release manner influenced at multiple regulatory levels by MMP-2.

Central to our model is the role of MMP-2 in activation of latent TGF-β1 in chemotherapy-treated stromal cells. We have previously determined that bone marrow stromal cells used in our model predominantly express high levels of MMP-2 (data not shown). In the current study, we demonstrated that MMP-2 acted as an activator of TGF-β1 in human bone marrow stromal cells and was required for optimal Smad3 phosphorylation after etoposide exposure. The rationale for focusing on MMP-2 was based on the observation that MMP-2 is secreted into ECM in association with remodeling during tissue injury and repair [42, 43] and our own data, which indicated that MMP-2 activity is rapidly elevated after etoposide treatment (Fig. 4A). MMP-2 activation paralleled phosphorylation of Smad3, occurring as early as 5 minutes after treatment. Inhibition of MMP-2 activity with OA-Hy diminished chemotherapy-induced phospho-Smad3 signals in stromal cells (Fig. 4B). These data suggested a critical role of MMP-2 in converting the ECM-bound latent TGF-β1 into its active form. Evaluation of MMP-2 knockout-derived stroma indicated a critical role for MMP-2 during activation of chemotherapy-induced TGF-β1/Smad3 signaling in bone marrow stromal cells (Figs. 4C, 4D). Transfection of human MMP-2 into murine stromal cells further suggests that MMP-2 plays a pivotal role in chemotherapy-induced activation of TGF-β1 in our model (Fig. 4E). Upregulation of MMP-2 and MMP-9 expression after TGF-β1 stimulation in several cell types has been well-documented [4447]. Thus, the current finding suggests that there is potentially a regulatory feedback loop in the bone marrow matrix during chemotherapeutic stress.

Evidence suggests that generation of ROS can serve as a secondary message to initiate signal transduction, in addition to its role in mediating apoptosis [48, 49]. In our model, we found that after chemotherapeutic stimulation, ROS was rapidly generated in bone marrow stromal cells (Fig. 5A). The drugs that induced phosphorylation of Smad3 were those that also induced ROS generation. Reduction of intracellular ROS by NAC reduced the phospho-Smad3 level induced by etoposide (Fig. 5B).

These data suggest a connection between ROS and MMP-2 activity when combined with the evidence that in vitro activation of pro–MMP-2 by hydrogen peroxide induced a dose-dependent phosphorylation of stromal cell Smad3 (Fig. 5C). The connection is strengthened by the observation that ROS activation of TGF-β1 is dependent on the presence of MMP-2 (Fig. 5D). In contrast to a recent report in which latent TGF-β1 could be directly activated by asbestos-derived ROS in A549 and mink pulmonary epithelial cells [50], our data indicate that dependence of TGF-β1 activation on MMP-2 cannot be circumvented during chemotherapeutic stress in marrow stromal cells. Consistent with the reports [51, 52] in which MMP-2 was activated by ROS in other cell models, our results suggest that generation of ROS is an early event that initiates the TGF-β1 signaling pathway in bone marrow stromal cells during chemotherapy. Of note, Ara-C exposure induced phosphorylation of Smad3 in stromal cells but did not promote intracellular ROS production. This suggests that other mechanisms are responsible for activation of Smad3 during Ara-C treatment.

C-terminal phosphorylation by the type I receptor of TGF-β is considered a key event in Smad activation [7]; however, there is evidence indicating other kinase pathways may also regulate Smad signaling [53, 54]. In our stromal cell model, P38, but not Erk1/2 or JNK, modulated Smad3 phosphorylation after etoposide treatment (Fig. 6B). This indicates that p38 may serve as a signal transducer that is downstream of TGF-β1/receptor ligation and directly mediates phosphorylation of Smad3. Phosphorylation of Smad3 by TGF-β receptor I in its C-terminus, or by p38 in its joint region [55], may initiate distinct signaling and induce different biological consequences.

Translocation of Smad3 from the cytoplasm into the nucleus of stromal cells treated with etoposide is the hallmark of activation of TGF-β1 signaling (Fig. 7). As a transcriptional regulator, nuclear Smads target a variety of genes [6, 7, 20]. Interestingly, a recent report documented that TGF-β1 stimulation led to downregulation of SDF-1 expression in bone marrow MS-5 stromal cells, although it was not investigated whether this was a Smad3-dependent effect [56]. It has also been shown that TGF-β1–treated stromal cells have less cell-surface VCAM-1 expression [57]. These reports are consistent with our earlier findings that etoposide-treated stromal cells have diminished chemotactic support [22] and impaired VCAM-1 expression [38] and our recent data indicating that these same cells, when treated with rh-TGF-β1, have impaired support of pro-B cell adhesion and proliferation (Figs. 8B, 8C).

Our current model suggests that dose-escalated chemotherapy may initiate a ROS/MMP-2–dependent activation of TGF-β1, which may have direct influence on hematopoietic cells as well as effects on stromal cell gene expression. Further investigation of these chemotherapy-induced changes may lend insight into strategies to protect the hematopoietic microenvironment during treatment in order to enhance hematopoietic recovery. Of note, very distinct pathways may be initiated during the acute and chronic phases of the stress response. Consequently, conclusions regarding the effects of chemotherapy exposure must be interpreted within the appropriate context as we attempt to better understand the dynamic response of the bone marrow to chemotherapy.


This work was supported by NIH grant R01 HL056888 (to L.F.G.) and NIEHS Training grant ES010953 (to S.C.). The authors would like to acknowledge Dr. Michael Reiss (The Cancer Institute of New Jersey, New Brunswick, NJ), who generously provided phospho-Smad2 antibody used in experiments that preceded those shown in the current report as well as anti-phospho-Smad3 antibody before commercial availability.