Pivotal Role of Connective Tissue Growth Factor in Lung Fibrosis

Transcription

1 ARTHRITIS & RHEUMATISM Vol. 60, No. 7, July 2009, pp DOI /art , American College of Rheumatology Pivotal Role of Connective Tissue Growth Factor in Lung Fibrosis MAPK-Dependent Transcriptional Activation of Type I Collagen Markella Ponticos, 1 Alan M. Holmes, 1 Xu Shi-wen, 1 Patricia Leoni, 1 Korsa Khan, 1 Vineeth S. Rajkumar, 1 Rachel K. Hoyles, 1 George Bou-Gharios, 2 Carol M. Black, 1 Christopher P. Denton, 1 David J. Abraham, 1 Andrew Leask, 3 and Gisela E. Lindahl 1 Objective. Connective tissue growth factor (CTGF; CCN2) is overexpressed in systemic sclerosis (SSc) and has been hypothesized to be a key mediator of the pulmonary fibrosis frequently observed in this disease. CTGF is induced by transforming growth factor (TGF ) and is a mediator of some profibrotic effects of TGF in vitro. This study was undertaken to investigate the role of CTGF in enhanced expression of type I collagen in bleomycin-induced lung fibrosis, and to delineate the mechanisms of action underlying the effects of CTGF on Col1a2 (collagen gene type I 2) in this mouse model and in human pulmonary fibroblasts. Methods. Transgenic mice that were carrying luciferase and -galactosidase reporter genes driven by the Col1a2 enhancer/promoter and the CTGF promoter, Supported by grants from the Arthritis Research Campaign, the Raynaud s and Scleroderma Association, the Welton Foundation, the Scleroderma Society, the Ontario Thoracic Society, the Canadian Institutes of Health Research, and the Canada Foundation for Innovation. Dr. Leask is recipient of an Early Researcher Award from the Ontario Ministry of Research and Innovation and is an Arthritis Society of Canada (Scleroderma Society of Ontario) New Investigator. 1 Markella Ponticos, PhD, Alan M. Holmes, PhD (current address: Novartis Horsham Research Centre, Horsham, UK), Xu Shi-wen, PhD, Patricia Leoni, PhD, Korsa Khan, BSc, Vineeth S. Rajkumar, PhD, Rachel K. Hoyles, MRCP, Carol M. Black, MD, FRCP, DBE, Christopher P. Denton, PhD, FRCP, David J. Abraham, PhD, Gisela E. Lindahl, PhD: Royal Free and University College Medical School, London, UK; 2 George Bou-Gharios, BSc, PhD: Imperial College School of Medicine, London, UK; 3 Andrew Leask, PhD: University of Western Ontario, London, Ontario, Canada. Drs. Abraham, Leask, and Lindahl contributed equally to this work. Dr. Leask owns stock in FibroGen. Address correspondence and reprint requests to Andrew Leask, PhD, Division of Oral Biology, Schulich School of Medicine and Dentistry, University of Western Ontario, London, Ontario N6A 5C1, Canada. Andrew.Leask@schulich.uwo.ca. Submitted for publication June 12, 2008; accepted in revised form March 23, respectively, were injected with bleomycin to induce lung fibrosis (or saline as control), and the extracted pulmonary fibroblasts were incubated with CTGF blocking agents. In vitro, transient transfection, promoter/ reporter constructs, and electrophoretic mobility shift assays were used to determine the mechanisms of action of CTGF in pulmonary fibroblasts. Results. In the mouse lung tissue, CTGF expression and promoter activity peaked 1 week after bleomycin challenge, whereas type I collagen expression and Col1a2 promoter activity peaked 2 weeks postchallenge. Fibroblasts isolated from the mouse lungs 14 days after bleomycin treatment retained a profibrotic expression pattern, characterized by greatly elevated levels of type I collagen and CTGF protein and increased promoter activity. In vitro, inhibition of CTGF by specific small interfering RNA and neutralizing antibodies reduced the collagen protein expression and Col1a2 promoter activity. Moreover, in vivo, anti-ctgf antibodies applied after bleomycin challenge significantly reduced the Col1a2 promoter activity by 25%. The enhanced Col1a2 promoter activity in fibroblasts from bleomycintreated lungs was partly dependent on Smad signaling, whereas CTGF acted on the Col1a2 promoter by a mechanism that was independent of the Smad binding site, but was, instead, dependent on the ERK-1/2 and JNK MAPK pathways. The CTGF effect was mapped to the proximal promoter region surrounding the inverted CCAAT box, possibly involving CREB and c-jun. In human lung fibroblasts, the human COL1A2 promoter responded in a similar manner, and the mechanisms of action also involved ERK-1/2 and JNK signaling. Conclusion. Our results clearly define a direct profibrotic effect of CTGF and demonstrate its contribution to lung fibrosis through transcriptional activation 2142

2 TYPE I COLLAGEN ACTIVATION BY CTGF IN LUNG FIBROSIS 2143 of Col1a2. Blocking strategies revealed the signaling mechanisms involved. These findings show CTGF to be a rational target for therapy in fibrotic diseases such as SSc. Systemic sclerosis (SSc) is an autoimmune rheumatic disease that is characterized by severe multisystem connective tissue fibrosis, of which type I collagen is a major component. Although expression of the disease is heterogeneous, internal organ fibrosis occurs in the majority of patients with SSc. Thus, 40% of patients with diffuse SSc experience pulmonary fibrosis, a major cause of mortality (1). Elucidation of the pivotal mediators or key signaling pathways that are overactive in fibrosis is a prerequisite for designing better therapies for SSc and related disorders. Fibroblasts, the major regulator cells of collagen turnover in adult connective tissue, express little or no connective tissue growth factor (CTGF; CCN2), unless it is induced during, for example, wound healing, by mediators such as transforming growth factor (TGF ) or thrombin (2 5). However, CTGF is consistently overexpressed in fibrotic lesions in patients with SSc and in those with other connective tissue diseases, affecting organs including the dermis, liver, kidney, heart, and lung (6,7). Elevated CTGF expression also correlates with, and contributes to, expression of -smooth muscle actin ( -SMA) and the myofibroblast phenotype in, for example, repair and development of connective tissue (8,9). Therefore, CTGF has been proposed to be a potential key profibrotic mediator, and indeed, CTGF expression levels were shown to be an excellent surrogate marker for the severity of fibrosis in SSc (10). In strong support of this concept in the clinical setting, we recently reported findings from a functional genetic study in which a common sequence variant in the CTGF promoter was genetically associated with SSc, particularly in patients with the antitopoisomerase autoantibody specificity and fibrosing alveolitis (11). The disease-associated allele was demonstrated to have higher transcriptional activity, providing the first direct causative evidence that CTGF is involved in susceptibility to lung fibrosis in SSc (11). In vitro, CTGF has been demonstrated to play an important role in fibrogenic responses, in that this matricellular protein has been attributed a strong modulating effect on the proliferation of fibroblasts and the production, adhesion, and contraction of extracellular matrix (ECM) (12). CTGF acts through integrins and proteoglycans as an independent proadhesive molecule, and modifies adhesive responses to ECM and growth factors (9,13,14). Expression of CTGF is strongly induced by TGF in fibroblasts (2,3,5) and is a mediator of the profibrotic effects of TGF in this cell type (14). However, elevated CTGF expression in fibrotic fibroblasts cultured from the skin of scleroderma patients is independent of the autocrine actions of TGF and Smad signaling, but is dependent on Sp-1 activity and the presence of endothelin 1 (15,16). Indeed, it appears that in vivo and in vitro, CTGF is a cofactor required to enhance the fibrotic responses to TGF (14,17). One of the most common in vivo models used for studying fibrotic processes in the lung is the mouse model of pulmonary fibrosis occurring in bleomycininduced acute lung injury, also known to be TGF dependent (18,19). Binding of TGF to the TGF receptor complex induces phosphorylation of Smad3, which complexes with Smad4, translocates to the nucleus, and induces transcriptional responses by binding to Smad binding elements (SBEs) in target promoters (20). The fibroblasts present in bleomycin-induced fibrotic lesions show activated Smad3 (21), and bleomycin-induced lung fibrosis is reversed by application of an adenovirus encoding Smad7 (22) and is known to be less severe in Smad3-deficient mice (23). Increased CTGF expression has been observed in this mouse model (4,6). However, the contribution of CTGF to bleomycin-induced pulmonary fibrosis and the mechanisms underlying the action of CTGF have yet to be elucidated. Herein, we used the bleomycin model of lung fibrosis to assess the potential contribution of CTGF to fibrogenesis. Our results provide new insights into how CTGF regulates the expression of Col1a2 (collagen gene type I 2) in this model and in human lung fibroblasts (COL1A2), making our data potentially important in understanding the processes involved in SSc-associated pulmonary fibrosis. MATERIALS AND METHODS Mouse model and tissue processing. The 2 transgenic reporter mouse lines used in this study (Col1a2- and CTGFtransgenic mice), both of which were generated on the C57BL/6 background, have been described elsewhere and faithfully recapitulate the expression patterns of the endogenous genes (24,25). Col1a2-transgenic mice harbor a 17-kb fragment encompassing the proximal promoter and the farupstream tissue-specific enhancer region driving both the luciferase and the -galactosidase ( -gal) reporter genes (Luc and LacZ, respectively) (25). CTGF-transgenic mice (a kind gift from Dr. Gary Grotendorst, University of Miami) harbor 1 kb of the human CTGF promoter driving the LacZ reporter gene (24). The mice (all females, ages 5 9 weeks) were anesthetized and injected with 1 dose of saline (100 l) or bleomycin (0.12 units in 50 l of sterile saline) intratracheally (25). Mice were killed at 7, 14, or 21 days after treatment.

3 2144 PONTICOS ET AL For CTGF antibody experiments, bleomycin- and saline-treated mice were injected with a neutralizing anti- CTGF specific antibody (24), administered intravenously into the lateral tail vein on day 1 (7.5 mg/kg), day 5 (5 mg/kg), and day 9 (2.5 mg/kg) after intratracheal challenge. As a control, some animals were injected with chicken IgY (ChIgY; Sigma, St. Louis, MO). The half-life of the pigy3 anti-ctgf antibody was determined to be 2.5 days in the mice, by measuring the antibody level in the circulation using enzyme-linked immunosorbent assay (results not shown). In the antibody experiments, there were 6 8 animals in each group, unless stated otherwise. Lungs were divided for cell explant cultures and for histologic, protein, or RNA analyses. For analysis of proteins, the mouse lung tissue was homogenized in 0.1M KCl and dissolved in 0.02M Tris HCl, ph 7.4. Hydroxyproline (HYP) measurements, eosinophil peroxidase (EPO) and myeloperoxidase (MPO) assays, luciferase assays, and sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE)/Western blotting protocols were carried out using these tissue homogenates. For histologic analysis, the lung samples were prepared for cryosections or were embedded in paraffin. For RNA extraction, the lung samples were homogenized in TRIzol reagent (Invitrogen, Carlsbad, CA). For reporter assays, the lung tissue homogenates were mixed with Passive Lysis Buffer (Promega Luciferase Assay kit; Promega, Madison, WI). Determination of inflammatory enzyme activity and HYP assays. Quantitation of EPO and MPO was performed as described previously (26). Briefly, the lung homogenates were centrifuged, and the resulting pellets were resuspended in 0.5% cetyltrimethylammonium chloride, and then rehomogenized and respun. The supernatants were analyzed for the activity levels of MPO and EPO by diluting cell extracts with EPO- or MPO-dilution buffer. Samples were mixed with EPOor MPO-substrate buffer, and were then incubated and stopped with EPO- or MPO-stop buffer. The optical densities were measured at 490 nm (for EPO) and 450 nm (for MPO). The collagen content was determined by measuring the HYP levels, as described previously (27). Briefly, the resin-isolated collagen fraction was acid hydrolyzed from homogenates at 110 C overnight, and then dissolved and separated from the resin and subjected to a colorimetric reaction. The samples were read for collagen content on a spectrophotometer at 560 nm, against a standard curve generated using trans-4-hydroxy- L-proline (Sigma). Histochemical staining. Sections of wax-embedded lung tissue (5 m) were stained with hematoxylin and eosin or picro Sirius red reagents, in accordance with standard histologic procedures. Sections were also immunostained with -SMA specific antibodies (Sigma) and type I collagen specific antibodies (Novotec, St. Martin la Garenne, France), using standard immunohistochemical protocols. Specific collagen and -SMA staining was quantified in whole lung sections by image capture on an Axioskop Mot Plus microscope and analyzed using KS300 image-analysis software (Carl Zeiss Instruments, Thornwood, NY). For detection of transgene activity, the lung sections were stained for -gal according to the method described by Ponticos et al (25). Assessment of luciferase and -gal activities, and analysis of messenger RNA (mrna) levels. The activities of the reporter genes, -gal and luciferase, in tissue homogenates or cell lysates were measured using the Dual-Light chemiluminescence reporter gene assay system (Tropix, Bedford, MA) according to the manufacturer s instructions. The levels were normalized to the levels of total protein. In transient transfection assays, transfection efficiency was controlled by cotransfection with either an RSV-Luc or a CMV- -gal vector, depending on the reporter gene of the experimental vector. Total RNA was extracted from the lung samples with the use of TRIzol reagent, following the manufacturer s instructions. Relative levels of mrna for CTGF, Col1a2, and GAPDH were determined by Northern blot analysis, as described previously (28). Primary lung fibroblast cultures, and treatment with anti-ctgf antibodies and small interfering RNA (sirna) in vitro. Primary pulmonary fibroblasts were isolated by explant culture from the lungs of mice 14 days after treatment with either bleomycin or saline. The cells were maintained in Dulbecco s modified Eagle s medium (DMEM; Gibco BRL, Grand Island, NY) supplemented with 10% fetal calf serum (FCS), 100 units/ml penicillin, and 100 g/ml streptomycin, and cultured in a humidified atmosphere of 5% CO 2 in air. Fibroblasts were subcultured weekly at confluence and used between passages 2 and 5. To study the effect of the anti- CTGF antibody, along with stimulation with TGF, on collagen synthesis, fibroblasts were grown to near confluence in DMEM with 10% FCS, and transferred to DMEM with 0.1% bovine serum albumin for 24 hours. TGF 1 (4 ng/ml), U0126 (MEK-1/2 inhibitor, 10 M), SB (p38 inhibitor, 0.6 M), SP (JNK inhibitor, 40 M), 0.1% DMSO, anti- CTGF antibody (0.4 g/ml) (24), ChIgY as a control (0.4 mg/ml), or recombinant CTGF (13) were added to the cell cultures for a further 24 or 48 hours. Human primary fetal lung fibroblasts (HFL-1 cells) were obtained from American Type Culture Collection (Manassas, VA), and, as normal controls, human fibroblasts were obtained from the unaffected lungs of adult patients undergoing cancer-related resection surgery. Control fibroblasts were used after written consent was obtained and ethics approval was provided (28). The human fibroblasts, after being isolated in explant culture, were grown under similar conditions as those used for the mouse fibroblasts. The fibroblast cultures were then assessed in transient transfection assays. A mouse-specific CTGF sirna (CAATTACAGTAG- CACATTAATT) and a nontargeting negative control (AllStar Negative Control; Qiagen, Crawley, UK) were transfected into explanted saline-treated or bleomycin-treated lung fibroblasts using the TransIT-TKO transfection reagent (MIR2150; Mirus, Madison, WI) according to the manufacturer s instructions. The CTGF sirna was transfected at 2 concentrations (5 nm and 20 nm). The control sirna was used at 20 nm. Western blot analysis. Cells were washed with phosphate buffered saline and solubilized with Laemmli sample buffer. For analysis of collagen, the medium was removed and adjusted to 20% (volume/volume) ammonium sulfate, followed by incubation at 4 C overnight. The samples were centrifuged and the pellet was resuspended in Laemmli sample buffer with -mercaptoethanol for SDS-PAGE, followed by Western blotting. The antibodies used for immunodetection in the medium precipitates and cell or tissue lysates were rabbit anti-ctgf specific antibody (Santa Cruz Biotechnology, Santa Cruz, CA) and anti pro 1 collagen antibody (Southern Biotechnology, Birmingham, AL), or anti type I collagen antibody (Novotec), anti-gapdh (Abcam, Cambridge, UK), or anti-actin antibody (Santa Cruz Biotechnology).

4 TYPE I COLLAGEN ACTIVATION BY CTGF IN LUNG FIBROSIS 2145 Transient transfections with promoter/reporter and expression constructs. Fibroblasts were transfected according to the manufacturer s recommendations (FuGene; Roche, Indianapolis, IN) with collagen promoter/ -gal reporter constructs (human COL1A2 minimal promoter construct [hmp] obtained from Prof. Francesco Ramirez, New York University, and mouse promoter construct obtained from Prof. Benoit de Crombrugghe, University of Texas, Houston) and a control cytomegalovirus (CMV) promoter/renilla luciferase reporter construct (Clontech, Palo Alto, CA). The CTGF promoter/ reporter construct used, p800ctgf-seap, was a construct in which expression of a secreted alkaline phosphatase (SEAP) reporter was controlled by the CTGF promoter, as described previously (16). Expression values were obtained following the kit manufacturer s instructions (Tropix). Overexpression of CTGF was achieved using a construct in which expression of the CTGF gene was controlled by the CMV promoter (pcmv-ctgf, consisting of full-length CTGF complementary DNA cloned into pcdna3.1 [Invitrogen]; kindly provided by Dr. Gary Grotendorst, University of Miami, Florida). The Smad7 expression vector used (pflag-msmad7; herein referred to as cmvsmad7) was a kind gift from Prof. Joan Massague (Sloan-Kettering, New York, New York). As a control, an empty expression vector (pcmv-ev) was transfected into cells. Data are expressed as the mean SEM results from 1 of 3 independent experiments. Statistical analysis was performed using Student s t-test. Electrophoretic mobility shift assay (EMSA). Nuclear extracts from cells transfected with either the pcmv-ctgf expression vector or the pcmv-ev empty control vector were prepared as described previously (29). Binding reactions were prepared in EMSA binding buffer (Promega) using 2.5 g nuclear extract and a 5 -end 32 P- ATP labeled doublestranded oligonucleotide probe, used on its own or together with either a cold competitor consensus oligonucleotide (Promega or Santa Cruz Biotechnology) or a specific supershift/ blocking antibody (Santa Cruz Biotechnology). Reaction mixtures were incubated for 30 minutes at room temperature, and DNA/protein complexes were separated using PAGE. (For probe sequences, see Figure 5B.) Statistical analysis. Results are expressed as the mean SEM. The data sets were tested using the Anderson-Darling normality test (Minitab, State College, PA), and the data were found to have a normal distribution. Differences between groups were determined using Student s t-test (Excel; Microsoft, Redmond, WA). P values less than 0.05 were considered significant. RESULTS Differential kinetics of CTGF and Col1a2 promoter/reporter transgene expression in vivo. To assess the role of CTGF in Col1a2 expression, we used both CTGF promoter/reporter transgenic and Col1a2 promoter/reporter transgenic mice and monitored the parameters of inflammation and fibrosis at 7, 14, and 21 days after intratracheal administration of a single dose of bleomycin (Figure 1). MPO and EPO levels were measured in lung lysates to assess inflammatory responses. Levels of both enzymes reached a maximum at 7 days after treatment (Figure 1A). Collagen content, as assessed by Figure 1. Characterization of bleomycin-induced fibrosis in lungs from promoter/reporter-transgenic mice and explanted lung fibroblasts. On days 0, 7, 14, or 21 after bleomycin or saline challenge, the lungs were removed and divided into tissue sections that were homogenized and assayed for various parameters. A, Bleomycin-induced fibrotic lungs were assessed for levels of myeloperoxidase (MPO), eosinophil peroxidase (EPO), and hydroxyproline (HYP) as well as expression of the Col1a2 promoter/reporter transgene (ColA2-luc-Tg) (measured as luciferase activity) and connective tissue growth factor (CTGF) promoter transgene (CTGF-LacZ-Tg) (measured as galactosidase activity). Bars on day 14 show the mean and SEM percentage increase relative to that in saline-treated lungs in 6 animals. Bars on days 7 and 21 show the mean and range of measurements from 2 mice. B, Northern blotting was used to assess the levels of endogenous CTGF and Col1a2 mrna in saline-treated (Sal) and bleomycintreated (Bleo) lung homogenates. GAPDH was used as a control. C, Histologic sections of lungs obtained 14 days after treatment with saline or bleomycin were assessed with hematoxylin and eosin (H&E) staining to visualize tissue architecture, with -smooth muscle actin ( -SMA) immunohistochemical staining to detect myofibroblast activity, and with -galactosidase ( -Gal) staining to demonstrate Col1a2 promoter/reporter activity (indicated by blue staining). D, Cellassociated expression of CTGF and levels of secreted type I collagen protein in explanted fibroblasts with or without 24 hours of stimulation with transforming growth factor (TGF ) were determined by Western blot analysis. Actin was used as a loading control. Fibroblasts explanted from lungs 14 days post bleomycin treatment retained their high-level expression of CTGF and collagen in vitro. Color figure can be viewed in the online issue, which is available at arthritisrheum.org.

5 2146 PONTICOS ET AL measuring the levels of HYP, was observed to increase over a period of 14 days after bleomycin treatment. In parallel, expression of the collagen promoter and CTGF promoter transgenes was measured in lung homogenates that were derived from the appropriate reporter-transgenic mice. A progressive increase in the activity of the collagen transgene (ColA2-luc-Tg) up to 14 days posttreatment, followed by a marked drop in activity by 21 days posttreatment, was observed (Figure 1A). These results mirrored the HYP analysis. In the CTGF promoter transgenic mice, the activity of the transgene (CTGF-LacZ-Tg) peaked at 7 days after treatment (Figure 1A), which was prior to the attainment of maximal expression of collagen but coincided with maximum EPO and MPO activity (Figure 1A). We also determined the levels of endogenous CTGF mrna and Col1a2 mrna in the Col1a2 promoter/reporter transgenic mice, using Northern blot analysis of whole lung extracts (Figure 1B). CTGF mrna levels peaked 7 days after bleomycin treatment, whereas Col1a2 mrna levels reached a maximum at 14 days post bleomycin challenge. Histologically, the lungs of mice obtained 14 days after bleomycin treatment were characteristically fibrotic in appearance, with disrupted alveolar architecture replaced by continuous cellular interstitial connective tissue (Figure 1C, top). Immunostaining using an -SMA specific antibody revealed positive loci within the fibrotic areas in bleomycin-challenged lungs, suggesting a presence of myofibroblasts (Figure 1C, middle). -gal staining of sections, to monitor Col1a2-driven reporter gene (LacZ) activity, revealed very low or no expression of the transgene in saline-treated control lungs but strong patchy staining in lungs with bleomycininduced fibrosis (Figure 1C, bottom). Taken together, these findings indicate that a robust induction of CTGF transcription occurs prior to enhanced collagen expression and synthesis in this model of lung fibrosis in vivo. Retention of enhanced collagen expression by explanted fibroblasts from bleomycin-treated lungs, and normalization of collagen levels in vitro by CTGF blockade. Molecular mechanisms underlying the activity observed were studied using mouse lung fibroblasts isolated from the fibrotic lesions 2 weeks after injection of bleomycin. Western blot analysis showed that there was a substantial increase in the levels of CTGF and type I collagen protein in fibroblasts explanted from bleomycin-challenged lungs, as compared with the levels in saline-treated control fibroblasts (Figure 1D, top). Both bleomycin- and saline-treated lung fibroblasts exhibited further enhancement in the levels of CTGF and type I collagen protein in response to TGF 1 (Figure 1D, bottom). These results indicate that fibroblasts explanted from mouse lungs with bleomycin-induced fibrosis retain their high-level collagen and CTGF expression, and that these fibroblasts can therefore be utilized to investigate the molecular mechanisms underlying the enhanced collagen expression in this model. The contribution of CTGF to the induced expression of Col1a2 in these cells was tested using an sirna oligonucleotide recognizing a mouse CTGF target sequence (Figure 2A). The sirna was transfected into the explanted fibroblasts at 2 concentrations (5 nm and 20 nm), and a nontargeting sirna (20 nm) was used as a control. The CTGF sirna, when used at both concentrations, efficiently blocked the enhancement of CTGF protein levels and resulted in inhibition of secreted collagen in fibroblasts from both the saline-treated and bleomycin-treated mouse lungs. In the presence of TGF 1, the CTGF sirna at the higher concentration was also able to block expression of CTGF and type I collagen protein in an efficient manner in saline-treated cells; however, in cells from bleomycin-treated mice, the reduction in collagen expression in the presence of sirna was often small. Figure 2A shows representative results from independent blotting experiments in which cells isolated from 3 mice from each group were assessed. Furthermore, we found that, in these explanted cells, addition of a neutralizing anti-ctgf antibody (pigy3) in vitro resulted in a significant decrease in secreted collagen production (Figure 2B). Cell preparations from bleomycin-induced fibrotic lungs were treated with either a specific CTGF antibody (pigy3) (24) or a chicken isotype control (ChIgY). The amount of secreted type I collagen protein was significantly decreased upon treatment with the anti-ctgf specific antibody, whereas cells treated with the ChIgY control showed no significant reduction. Addition of a panspecific anti-tgf antibody, 1D11, had a similar effect, confirming that collagen production in this model of lung fibrosis is dependent on the activity of TGF. In addition, in experiments using the neutralizing anti-ctgf antibody, we also observed a 40.5% reduction in Col1a2 transgene activity in cell preparations from 3 bleomycin-treated lungs (mean SEM relative light units [RLU] in pigy3-treated lung fibroblasts versus RLU in untreated controls; P 0.001) (Figure 2C, top), suggesting that there is regulatory activity taking place at the transcriptional level.

6 TYPE I COLLAGEN ACTIVATION BY CTGF IN LUNG FIBROSIS 2147 Figure 2. Effects of connective tissue growth factor (CTGF) blockade on type I collagen expression in fibroblasts explanted from bleomycin- or saline-treated mouse lungs (at day 14 posttreatment), and confirmation of the effects in vivo. A, Mouse-specific CTGF small interfering RNA (sirna) (5 nm and 20 nm) or control sirna (20 nm) was transfected into bleomycin- or saline-treated mouse lung fibroblasts, and CTGF and type I collagen protein levels, with or without stimulation with transforming growth factor (TGF ), were determined by Western blot analysis after 48 hours. GAPDH was used as a loading control. B, Explanted fibroblasts from saline- or bleomycin-challenged lungs were left untreated or treated for 72 hours with anti-ctgf antibody (pigy3), anti-tgf antibody (1D11), or control antiserum (chicken IgY [ChIgY] or mouse IgG), and type I collagen expression was assessed by Western blotting. Actin was used as a loading control. C, Fibroblasts explanted from bleomycin-challenged lungs from 3 animals per group were treated with anti-ctgf pigy3 or control ChIgY antibody, and expression of the endogenous Col1a2 promoter transgene (Col1A2-Luc-Tg) in the cell lysates was determined in vitro as the fold induction of luciferase activity. Expression of Col1A2-Luc-Tg transgene was measured in vivo at day 14 in saline- or bleomycin-treated (Bleo) mice (n 8 animals per group) left untreated or injected with either anti-ctgf pigy3 antibody or control ChIgY preimmune serum. P 0.02 versus ChIgY control lung samples. D, Bleomycin-treated lung samples were injected with either anti-ctgf pigy3 or control ChIgY, and whole lung sections were stained with picro Sirius Red or specific antibodies to type I collagen or -smooth muscle actin ( -SMA) to assess the in vivo effects on lung tissue architecture, collagen deposition, and -SMA expression, respectively. Bars 100. Results of immunohistochemical stainings are shown quantitatively, in anti-ctgf pigy3 or control ChIgY treated cells, as the area of the specific antibody staining for type I collagen and -SMA over the total stained area in 4 nonconsecutive whole lung sections from 3 animals per group. Bars in C and D show the mean and SEM. Contribution of CTGF to the fibrotic phenotype in bleomycin-injured lungs in vivo. To assess the contribution of CTGF to the induced expression of collagen in vivo, we administrated the specific anti-ctgf antibody into Col1a2 promoter/reporter transgenic mice after induction of fibrosis with bleomycin challenge. Since previous data showed that the maximum inflammatory response occurred at 7 days postchallenge and the collagen promoter activity peaked at 14 days after bleomycin challenge (Figure 1), we examined the effects of anti-ctgf treatment on these parameters on day 7 and day 14, respectively. Animals treated with the anti- CTGF antibody showed a modest, yet significant, decrease (26%) in the amount of collagen transgene activity (mean SEM RLU) relative to that in the isotype controls ( RLU; P 0.01) (Figure 2C, bottom). However, neither the anti-ctgf antibody treatment nor treatment with the isotype control affected the levels of MPO or EPO at day 7 (results not shown). Flow cytometry analyses of the number of neutrophils, macrophages, CD4-positive cells, and CD8- positive cells confirmed that the impact of the anti- CTGF antibody on the inflammatory response was negligible (results not shown). Thus, anti-ctgf antibody treatment in vivo appeared not to affect the immune response specifically, and yet resulted in a significant reduction in collagen promoter activity. To verify that the in vivo antibody treatment also affected the protein levels of fibrosis-related genes, we analyzed the lung tissue sections with specific staining for type I collagen and -SMA. Visual inspection showed reduced staining of these proteins, and also revealed an improvement in lung tissue architecture after anti-ctgf antibody treatment (Figure 2D). The reduction in antibody staining was quantitatively determined in 4 nonconsecutive whole lung sections from 3 animals in each group (Figure 2D, graphs), and it was found that there was an 82% reduction in type I collagen levels (P versus isotype controls) and a 73% reduction in -SMA staining (P versus isotype controls), which was probably not completely attributable to the decrease in collagen transcriptional activity, but also could be attributed to other effects of CTGF on profibrotic processes.

7 2148 PONTICOS ET AL Figure 3. Enhancement of Col1a2 promoter activity occurring independent of Smad signaling in fibroblasts from bleomycin-challenged mouse lungs. A, A 354-bp Col1a2 promoter/reporter construct (ColIa2 wild-type [WT]) or a construct bearing a mutated Smad binding element but otherwise identical to wild-type (ColIa2 smad) was transfected into fibroblasts derived from saline control or bleomycin-treated animals, and Col1a2 promoter activity was assessed as the fold induction of -galactosidase activity. B, Bleomycin-treated fibroblasts were transfected as described in A and treated with anti connective tissue growth factor (anti-ctgf) antibody pigy3 or chicken IgY (ChIgY) as control, and Col1a2 promoter activity was assessed. C, Saline- or bleomycin-treated fibroblasts were cotransfected with or without the Col1a2 promoter/reporter constructs and with or without the CTGF expression vector (cmvctgf) or empty control vector (cmvev), and Col1a2 promoter activity was assessed. D, A CTGF promoter/reporter construct, comprising a secreted alkaline phosphatase (SEAP) reporter controlled by the CTGF promoter, was cotransfected with or without an expression vector for Smad7 (cmvsmad7) into fibroblasts derived from saline- or bleomycin-treated mice, and the fold induction of CTGF promoter activity was assessed. Bars show the mean and SEM. NS not significant. Contribution of CTGF to Col1a2 promoter activity independent of Smad signaling. Having confirmed an effect of blocking CTGF in vivo, we again used explanted fibroblasts to further investigate the mechanisms underlying the enhanced Col1a2 transcription by CTGF. Since TGF -induced collagen expression has been shown to involve Smad signaling in this model in vivo (22,23), we tested whether the CTGF effect was dependent on this pathway in vitro. Indeed, transfection of a Col1a2 promoter/reporter construct bearing a mutated SBE into bleomycin-derived fibroblasts confirmed that Smad signaling was involved in bleomycin-induced procollagen transcription in our system. Loss of the SBE site resulted in an 50% reduction in Col1a2 promoter activity in bleomycin-challenged lung fibroblasts (mean SEM fold induction of -gal activity in wild-type cultures versus in mutated construct cultures; P 0.05) (Figure 3A). Nonetheless, a significant Smad-independent component to the elevated Col1a2 promoter activity remained (Figure 3A). The Col1a2 promoter activity in bleomycininjured lungs was further reduced by the anti-ctgf antibody in wild-type construct cultures (reduction in -gal activity 41.3% in pigy3-treated cultures; P 0.01 versus isotype controls) (Figure 3B), and the effect was independent of the mutated SBE (reduction in -gal activity 53.1% in pigy13-treated cells in mutated construct cultures; P 0.01 versus isotype controls), suggesting that CTGF acts on collagen transcription in a manner independent of Smad binding. Conversely, overexpression of CTGF that was induced by the pcmv- CTGF expression vector in saline-treated cells enhanced Col1a2 transcription 4-fold as compared with that in saline-treated cells cultured with the pcmv-ev empty control vector, and the effects were independent of the SBE mutation (Figure 3C). Interestingly, in bleomycin-treated cells, the enhanced activity of the wild-type promoter could not be further induced by CTGF overexpression. However, the reduced activity of the Smad-mutated Col1a2 construct in bleomycin-treated fibroblasts was enhanced to near the level of the wild-type promoter activity by induction of CTGF overexpression (increased from a mean SEM fold induction of -gal activity of to ; P 0.04 versus controls) (Figure 3C), again confirming an activation independent of the SBE. We also tested the effect of Smad signaling on CTGF transcription in this system. Cotransfecting a Smad7 expression vector (cmvsmad7) together with a CTGF promoter/reporter construct (p800ctgf-seap) reduced the bleomycin-induced promoter activity by 60% (mean SEM fold induction of CTGF promoter activity in cultures with p800ctgf-seap plus cmvsmad7 versus in cultures with

8 TYPE I COLLAGEN ACTIVATION BY CTGF IN LUNG FIBROSIS 2149 Figure 4. Effect of inhibition of MAPK pathways on Col1a2 promoter activity in fibroblasts from saline- and bleomycin-treated mouse lungs. A, Cells were transfected with the Col1a2 promoter construct and then treated with an ERK-1/2 inhibitor (U0126 [U0] at 10 M), a p38 inhibitor (SB [SB] at 6 M), a JNK inhibitor (SP [SP] at 40 M), or DMSO (0.1%) as vehicle control. Results are the mean and SEM percentage of Col1a2 promoter expression relative to the DMSO control. B, Cells cotransfected with the Col1a2 promoter/reporter and cmvctgf construct or empty control vector (cmvev) were incubated with or without the same MAPK inhibitors as described in A. Bars show the mean and SEM fold induction of -galactosidase activity. C, Effects of the MAPK inhibitors on levels of secreted type I collagen protein were assessed by Western blotting in cells treated with or without recombinant connective tissue growth factor (recctgf; 20 ng/ml). The sample loading was adjusted for cell number, and results are shown relative to the GAPDH protein level in the cell layer. p800ctgf-seap alone; P 0.02), but without any effect on the levels in saline-treated cells (Figure 3D). This indicates a strong contribution from this pathway to the expression of CTGF in this model of lung fibrosis, which is consistent with previously published observations in vivo (22,23). Our results suggest that the elevated level of CTGF present in fibroblasts cultured from fibrotic lungs contributes directly to the overproduction of type I collagen at the transcriptional level, and appears to mediate a significant part of bleomycininduced fibrotic processes. Involvement of the ERK-1/2 and JNK signaling pathways, but not p38, in CTGF-induced Col1a2 transcriptional activity. We next investigated the possible involvement of the 3 common MAPK pathways, ERK- 1/2, p38, and JNK (Figure 4A). Transient transfections in bleomycin- or saline-treated lung-derived fibroblasts, together with small molecule inhibitors for these 3 pathways at optimal concentrations, resulted in a significant reduction in Col1a2 promoter activity by the ERK-1/2 pathway inhibitor U0126 (10 M) (reduction of 44.4%; P versus DMSO control), but no effect was observed with inhibitors of the p38 pathway (SB at 0.6 M) or the JNK pathway (SP at 40 M) in saline-treated cells. In contrast, in bleomycintreated cells, the Col1a2 promoter activity was significantly reduced by both U0126 (reduction of 59%; P 0.01 versus DMSO control) and SP (reduction of 72%; P 0.05 versus DMSO control), but not by SB (Figure 4A). We therefore further investigated the involvement of the ERK-1/2 and JNK pathways on the CTGF effect in isolation (Figure 4B). We transiently cotransfected the mouse Col1a2 wild-type promoter/reporter construct together with the pcmv-ctgf vector, or with pcmv-ev as a control, into saline-treated cells, and subsequently treated the cells with inhibitors of these 2 pathways. In cells transfected with pcmv-ev, only U0126 inhibited Col1a2 promoter activity. However, in cells transfected with the CTGF expression vector, SP also significantly reduced the Col1a2 activity (by 30%; P 0.01 versus DMSO control), although U0126 had a greater effect (reduction of 66%; P versus DMSO control) (Figure 4B). This result mirrored the findings in cells from bleomycintreated lungs, suggesting a major role for CTGF in Col1a2 induction after bleomycin-induced lung injury, with mechanisms of action that are dependent on the activities of the ERK-1/2 and JNK pathways. The effect of the ERK-1/2 and JNK pathway inhibitors was also determined at the type I collagen protein level by treating fibroblasts derived from salinetreated control lungs with recombinant CTGF (20 ng/ ml) for 24 hours after the addition of the inhibitors (at the same concentrations as noted above). In these experiments, U0126 reduced the type I collagen protein levels below baseline levels, whereas, similar to the results in the transfection assays, the JNK inhibitor had less of a reducing effect (Figure 4C). Activation of the Col1a2 promoter by CTGF through a region containing the proximal inverted CCAAT element. To further delineate the transcriptional mechanism underlying the CTGF-induced collagen promoter activity, we cotransfected into fibroblasts pcmv-ctgf (or pcmv-ev) together with a promoter/reporter construct containing the minimal Col1a2 promoter inserted upstream of -gal (Col-354)

9 2150 PONTICOS ET AL Figure 5. Connective tissue growth factor (CTGF) activation of the mouse Col1a2 minimal promoter via sequences adjacent to a CCAAT element in mouse lung fibroblasts. A, A proximal promoter/reporter construct (Col-354), as well as 2 deletion constructs, one excluding the Smad response element (Col-96) and one with a further deletion also excluding the CCAAT box (Col-50), plus a full-length construct with a mutated CCAAT box (CCAATmut), were transfected into fibroblasts together with either the expression vector encoding CTGF (cmvctgf; indicated as ) or the empty expression vector (indicated as ). Bars show the mean SEM fold induction of -galactosidase activity. B, Electrophoretic mobility shift assay (EMSA) was performed to assess transcription factor binding. A Col1a2 CCAAT-box oligonucleotide (oligo) was radioactively labeled and used as a probe in EMSA with nuclear extracts from cells transfected with cmvctgf ( ) or empty control vector cmvev ( ). Various antibodies (Ab) and unlabeled oligonucleotides were used in supershift (SS)/blocking and competition reactions, respectively. Three complexes, apart from the well-characterized C-promoter binding factor (CBF) complex, were formed (denoted CI, CII, and CIII). The diagram on the left shows the probes used in EMSA, including the functional binding site for CBF and putative binding sites for CREB, c-jun, and Sp-1. C/EBP CCAAT box/enhancer binding protein ; AP-1 activator protein 1. and various deletion/mutation constructs thereof (Figure 5A). Deletion of a segment of DNA bearing the known TGF -responsive SBE (denoted CAGA, or Col- 96) and surrounding sequences still permitted induction of collagen activity by CTGF (by 4.5-fold; P versus empty vector control) (Figure 5A). However, a further deletion of a segment containing a previously characterized CCAAT box (denoted Col-50) abolished CTGF responsiveness. The critical involvement of the CCAAT box was confirmed by site-directed mutagenesis, a technique that has been previously shown to abolish function, i.e., binding of C-promoter binding factor (CBF)/nuclear factor Y (NF-Y) to this site (30), on the full-length construct. As a control, cells in parallel wells transfected with these constructs were treated with TGF 1 instead of pcmv-ctgf, and as expected, the promoter fragment containing the Smad response element was found to be necessary for TGF induction (results not shown). These results show that the proximal inverted CCAAT promoter motif is necessary for CTGF transcriptional induction, and that CTGF activates a fibrotic response by augmenting activity of collagen transcription via sequences around the proximal inverted CCAAT box. Association of CTGF overexpression with alteration of transcription factor binding at sequences spanning the proximal inverted CCAAT box, and possible involvement of CREB and c-jun. Since the proximal inverted CCAAT box, a known CBF/NF-Y binding site (30), is central to Col1a2 transcription, and its deletion leads to a significant reduction in basal Col1a2 promoter activity, we examined whether CTGF affects CBF binding activity directly, and/or whether surrounding sequences are also involved. Thus, we used EMSA with a DNA oligonucleotide containing the CCAAT box and surrounding sequences to investigate the interaction of nuclear proteins extracted from cells in which CTGF was overexpressed by transfection with pcmv-ctgf, in comparison with control extracts transfected with pcmv-ev. The sequence of the probe was analyzed in silico using TF BIND, a software program identifying putative transcription factor elements (31). Using this probe, possible Sp-1 and CREB/c-Jun elements were identified in addition to CBF (Figure 5B). Importantly, a commercial CBF consensus oligonucleotide, used as an unlabeled competitor, also was found to contain a putative CREB site in addition to the CCAAT box. EMSA analysis of lung fibroblasts transfected

10 TYPE I COLLAGEN ACTIVATION BY CTGF IN LUNG FIBROSIS 2151 with pcmv-ctgf or an empty control vector showed that, while there was no significant difference in CBF/ NF-Y binding at this site (confirmed by a specific CBF supershift antibody) in response to overexpression of CTGF, 3 other complexes (denoted CI, CII, and CIII) displayed differential binding (Figure 5B). While CI was relatively weak and more difficult to interpret, CII displayed a clearer pattern, with a strong binding in control extracts and abolished binding in response to overexpression of CTGF. Specific antibodies and unlabeled consensus oligonucleotides were used to identify factors forming the CII complex. The CBF oligonucleotide, which contains a putative CREB binding site, blocked binding of CII as well as CBF. Moreover, a CREB consensus oligonucleotide significantly reduced binding of protein to DNA, whereas an activator protein 1 (AP-1) oligonucleotide had no discernable effect. Specific antibodies to CBF and c-jun abolished the CBF and CII complexes, respectively. Interestingly, antibodies to CCAAT box/enhancer binding protein (C/EBP ) and Sp-1 had a neutralizing effect on CII formation, indicating a possible interaction of these factors with this proximal complex. The final complex, CIII, consisted of a diffuse DNA/protein complex that appeared to alter in the presence of C/EBP and c-jun antibodies. Thus, our data suggest a possible involvement of CREB and c-jun. Responsiveness of the human minimal COL1A2 promoter to CTGF activation in human lung fibroblasts. To determine the effect of CTGF on type I collagen transcription in the human context, we cotransfected human COL1A2 promoter deletion constructs together with the CTGF expression vector into human pulmonary fibroblasts (Figure 6). A sequence alignment of the human and mouse procollagen type I 2 promoters showed a high degree of homology in the proximal region (Figure 6A). The sequences of interest, around the CCAAT box, contained some base differences when the 2 species were compared, especially across the putative CREB binding site (Figure 6A, as compared with Figure 5B). Interestingly, when the human sequence was analyzed in silico using TF BIND (31), a putative weak CREB site was identified, despite these base differences, suggesting that this site is functionally conserved. In transient transfection assays using human fetal lung fibroblasts (HFL-1 cells), a human COL1A2 minimal promoter construct (hmp) that was equivalent to mouse Col-354, a deletion of the region containing the Smad response element (Bst XI), and a deletion equivalent to mouse Col-96 (Sma I), all of which contained sequences including the CCAAT box and adjacent se- Figure 6. Proximal promoter sequences including the CCAAT box in the human COL1A2 promoter, and response to cmvctgf-induced overexpression of connective tissue growth factor (CTGF) in human lung fibroblasts. A, Sequence alignment of the human and mouse collagen gene type I 2 proximal promoters, showing sequence homology, published functional response elements, and deletion construct 5 ends. The restriction enzyme sites Bst XI and Sma I were used to generate the human deletion set. B, Diagram of the COL1A2 promoter/reporter deletion set, with the Smad (caga) and C-promoter binding factor (CBF) (ccaat) response elements indicated. C, Human COL1A2 promoter/ reporter activity in human fetal lung fibroblasts (HFL-1 cells) after transfection with cmvctgf ( ) or empty vector control ( ). D, Human COL1A2 promoter activity in cotransfection assays with normal control human adult lung fibroblasts. The fibroblasts were cotransfected with a human minimal promoter construct and the cmvctgf construct or empty control vector cmvev in the presence of MAPK pathway inhibitors U0126 (U0; 10 M) or SP (SP; 40 M) or both; DMSO was used as the control. Bars show the mean and SEM fold induction of COL1A2 promoter activity. C/EBP CCAAT box/enhancer binding protein ; AP-1 activator protein 1.

11 2152 PONTICOS ET AL quences (Figure 6B), were used. In all cultures of HFL-1 cells with these constructs, the COL1A2 promoter activity responded to CTGF overexpression (Figure 6C). Furthermore, we tested the effect of MAPK inhibitors on CTGF-induced activation of the human COL1A2 minimal promoter in control human adult lung fibroblasts, and observed a result very similar to that in the mouse system. The activity of COL1A2 in U0126- inhibited cells was reduced by 48% (P 0.02 versus DMSO control), while cells inhibited with SP reduced the COL1A2 activity by 30% (P 0.05 versus DMSO control), and when cells were treated with both together, the activity was reduced by 65% (P 0.01 versus DMSO control), suggesting that the ERK-1/2 and JNK pathways play important roles in this process (Figure 6D, as compared with Figure 4B). Again, more detailed studies, not within the scope of this report, are under way to delineate the precise molecular interactions involved in CTGF induction of COL1A2. DISCUSSION In this study, we used the Col1a2 promoter/ reporter transgenic mouse to assess whether CTGF affects type I collagen expression in bleomycin-induced lung fibrosis in vivo. We also used in vitro approaches to elucidate the possible mechanisms underlying the action of CTGF in mouse and human lung fibroblasts. Consistent with previous reports showing elevated CTGF expression in the bleomycin model of lung fibrosis (4,6), CTGF was expressed prior to collagen production and ECM accumulation, providing the possibility that CTGF is a key mediator in pulmonary fibrotic processes. Evidence for a direct involvement of CTGF in fibrosis has been provided by circumstantial observations. First, in a bleomycin-resistant mouse model (BALB/c), CTGF was not up-regulated in response to challenge with bleomycin (6). Moreover, Bonniaud and colleagues further demonstrated that adenovirus-based delivery of CTGF to a bleomycin-sensitive mouse strain (C57BL/6) caused a transient fibrotic response (32), but in the resistant BALB/c strain, the lung was rendered susceptible to bleomycin-induced fibrosis (33). We have built upon these prior findings by addressing whether CTGF affects lung type I collagen expression and Col1a2 promoter activity directly in vitro and in vivo. We did this, first, by demonstrating that fibroblasts explanted from fibrotic lungs after bleomycin treatment retained their high-level expression of CTGF and type I collagen in culture. Two approaches for antagonizing CTGF action were then utilized. Use of CTGF sirna reduced the expression of CTGF and type I collagen protein in bleomycin-challenged fibroblasts to that observed in saline-treated control lungs, and the sirna were able to inhibit TGF 1-induced overexpression in these cells. Furthermore, a neutralizing anti-ctgf antibody, previously used to block the CTGF effect on cell migration and fibronectin production in kidney mesangial cells (24), also reduced type I collagen expression and Col1a2 promoter activity in fibroblasts isolated from bleomycinchallenged lungs. Extending these results, the anti-ctgf antibody reduced the activity of the Col1a2 promoter/enhancerdriven transgene (Luc) in bleomycin-challenged lungs, also in vivo. Moreover, mice treated with the anti-ctgf antibody after bleomycin challenge displayed improved lung tissue architecture and reduced expression of type I collagen and -SMA. Thus, these data are consistent with recent findings showing that, whereas CTGF overexpression is, on its own, insufficient for causing lung fibrosis in vivo, CTGF expression produces a susceptibility to lung fibrosis and enhances the fibrotic response (32,33). The findings from a very recent report, in which application of the anti-ctgf antibody reduced TGF induced skin fibrosis in mice (34), also support our findings and strengthen the notion of a direct role for CTGF in fibrosis enhancement and progression. Taken together, these results strongly suggest that inhibition of CTGF may be of therapeutic value in profibrotic clinical settings. In this study, we also investigated the mechanisms underlying the ability of CTGF to induce Col1a2, by exploring which signaling and transcriptional regulatory pathways are involved. TGF is known to induce a robust increase in Col1a2 transcription by both Smaddependent and -independent mechanisms (35); however, type I collagen overexpression in fibrosis is less well-defined in terms of its signaling pathways. In accordance with studies in Smad3-knockout mice, in which it was shown that fibrosis and collagen expression were reduced in this bleomycin model of lung fibrosis (23), we found that the bleomycin-inducing effect on the Col1a2 promoter was partly Smad-dependent. However, the contribution from CTGF on Col1a2 transcription appeared not to require Smad binding, since the anti- CTGF antibody was able to reduce the activity of a Col1a2 promoter lacking its functional SBE, suggesting that there is a distinct means of regulation. We next investigated the possible involvement of the major MAPK pathways that have been implicated in fibrosis, notably in SSc-related lung fibrosis (36), and in CTGF regulation (37) and action (38). Our results showed that bleomycin-induced Col1a2 transcription