Postnatal induction of transforming growth factor β signaling in fibroblasts of mice recapitulates clinical, histologic, and biochemical features of scleroderma

Abstract

Objective

Increased signaling by transforming growth factor β (TGFβ) has been implicated in systemic sclerosis (SSc; scleroderma), a complex disorder of connective tissues characterized by excessive accumulation of collagen and other extracellular matrix components in systemic organs. To directly assess the effect of sustained TGFβ signaling in SSc, we established a novel mouse model in which the TGFβ signaling pathway is activated in fibroblasts postnatally.

Methods

The mice we used (termed TBR1^CA; Cre-ER mice) harbor both the DNA for an inducible constitutively active TGFβ receptor I (TGFβRI) mutation, which has been targeted to the ROSA locus, and a Cre-ER transgene that is driven by a fibroblast-specific promoter. Administration of 4-hydroxytamoxifen 2 weeks after birth activates the expression of constitutively active TGFβRI.

Results

These mice recapitulated clinical, histologic, and biochemical features of human SSc, showing pronounced and generalized fibrosis of the dermis, thinner epidermis, loss of hair follicles, and fibrotic thickening of small blood vessel walls in the lung and kidney. Primary skin fibroblasts from these mice showed elevated expression of downstream TGFβ targets, reproducing the hallmark biochemical phenotype of explanted SSc dermal fibroblasts. The mouse fibroblasts also showed elevated basal expression of the TGFβ-regulated promoters plasminogen activator inhibitor 1 and 3TP, increased Smad2/3 phosphorylation, and enhanced myofibroblast differentiation.

Conclusion

Constitutive activation of TGFβ signaling in fibroblastic cells of mice after birth caused a marked fibrotic phenotype characteristic of SSc. These mice should be excellent models with which to test therapies aimed at correcting excessive TGFβ signaling in human scleroderma.

Scleroderma, which is also known as systemic sclerosis (SSc), is a multisystem connective tissue disorder in humans that leads to fibrosis of the skin and internal organs and often results in death (1, 2). The etiology of SSc remains incompletely understood (3). Early pathologic hallmarks of the disease include immunologic activation associated with the development of a range of autoantibodies, microvascular endothelial cell activation, and later perivascular infiltration of mononuclear immune cells (4). As a consequence of these early events, fibroblasts within the skin and affected internal organs are activated, leading to excessive deposition of extracellular matrix (ECM) components (4). A proportion of fibroblasts differentiate into myofibroblasts, which are believed to be central to the development of fibrosis (5–7). As the disease becomes established, the skin becomes fibrotic and small blood vessel walls thicken. Eventually, the lungs, heart, kidneys, and gastrointestinal tract are involved, but there is substantial heterogeneity in the morbidity and mortality due to SSc (3). Secondary events, including exposure to environmental agents, hemodynamic stress, or epithelial injury, may trigger involvement of target organs. The disease can therefore be characterized as ubiquitous fibrosis of the skin and vasculature, together with increased susceptibility to more widespread organ-specific fibrosis.

There is considerable evidence implicating overactivity of the pleiotropic growth factor transforming growth factor β (TGFβ) in the pathogenesis of SSc (8, 9). Thus, elevated levels of TGFβ have been reported in SSc, and elevated expression of TGFβ protein or messenger RNA in the skin and lung has been reported (10). In addition, the altered profile of gene and protein expression characteristic of SSc fibroblasts is reminiscent of that of normal fibroblasts activated by TGFβ (11–13). It has therefore been hypothesized that TGFβ released from a variety of cell types activates fibroblasts in SSc. Testing this hypothesis experimentally is difficult owing to the pleiotropic effects of TGFβ on multiple cell types, its key role in embryonic development, and the complexity of TGFβ regulatory mechanisms in vivo.

Because of the proposed central role of TGFβ in SSc (8, 9) and the involvement of fibroblasts and fibroblast-like cells in the disease, we postulated that a transgenic mouse model in which TGFβ signaling in these cells is increased would recapitulate many key features of the disease. Investigators in our group had previously identified a strong transcriptional enhancer located between 15 kb and 19 kb upstream of the transcription start site of the Col1a2 gene that was largely fibroblast specific (14), so we attempted to generate transgenic mice in which this enhancer directed the expression of constitutively active type I TGFβ receptor (TGFβRI). Unfortunately, all embryos died early during gestation, presumably because there was a disruption of TGFβ signaling that caused embryonic death (data not shown). To circumvent this, we used the Cre/loxP system in which tamoxifen-inducible recombinase can be activated after birth. We were able to produce a mouse in which TGFβ signaling was disrupted only after the developmental progress of the embryo had been largely completed. We report here that the expression of constitutively active TGFβRI after birth in fibroblastic cells from these mice caused generalized skin fibrosis and a thickening of the walls of the small arteries, reproducing 2 of the hallmarks of human SSc.

MATERIALS AND METHODS

Generation of mice expressing constitutively active TGFβRI and Col1a2 Cre-ER fusion protein (TBR1^CA; Cre-ER mice).

Constitutively active TGFβRI complementary DNA (cDNA) was a gift from Dr. Joan Massague (15), and the wild-type (WT) ROSA26 targeting vector was from Dr. Phillip Soriano (16). Constitutively active TGFβRI cDNA was cloned in the vector pAB-P2 that contained the loxP sites flanking a pGK-neo-STOP cassette, the splice acceptor, and the 3′ and 5′ arms. The resulting vector pAB-T204D-P2 was cloned in the ROSA26-1MP vector. Both 5′ and 3′ external probes were used in Southern blots to detect the homologous recombinant. The targeted clone yielded a 2-kb fragment (instead of 11 kb) when digested with Eco RI and hybridized with the 5′ probe, and a 10-kb fragment (instead of 15 kb) when digested with Eco RV and hybridized with the 3′ probe.

The targeting vector was electroporated into AB1 cells that were derived from 129/SvEvTac mice. Neomycin-resistant clones were screened for homologous recombination at the ROSA locus. Male chimeras were generated by injecting targeted embryonic stem (ES) cells into C57BL/6 blastocysts and bred to WT females to confer germline transmission of the targeted allele. Heterozygous male progeny carrying the targeted allele were crossed with Col1a2 Cre-ER–transgenic females expressing the Cre-ER fusion protein in fibroblasts. Tail DNA from offspring was examined by polymerase chain reaction to confirm the presence of the knockin allele.

Histologic analysis.

Tissue samples were fixed in 10% buffered formalin and embedded in paraffin. Sections (5–6 μm) were stained either with hematoxylin and eosin or with Masson's trichrome. Immunofluorescence was performed on fibroblasts derived from the skin of mice that were injected with oil or with 4-hydroxytamoxifen (4-OHT). Antihemagglutinin (HA.11; Covance, Berkeley, CA) and anti–α-smooth muscle actin (anti–α-SMA; Sigma, St. Louis, MO) antibodies were used at 1:2,000 and 1:1,000 dilutions, respectively. Goat anti-mouse Alexa Fluor 594 (Molecular Probes, Eugene, OR) was used as secondary antibody. Image stacks were collected on a Nikon 2000U inverted microscope (Nikon, Tokyo, Japan) controlled by MetaMorph software (Universal Imaging, West Chester, PA) and deconvolved using AutoQuant AutoDeblur software (Media Cybernetics, Silver Spring, MD).

Immunohistochemistry was performed on paraffin-embedded sections by established methods. Primary antibodies to types I and III collagen (Santa Cruz Biotechnology, Santa Cruz, CA) were used at 1:100 and 1:500 dilutions, respectively. Immunohistochemistry with rabbit anti-human von Willebrand factor (vWF) antibody (Santa Cruz Biotechnology) was performed at a dilution of 1:100. The slides were incubated with biotin-conjugated secondary antibody (1:200; Vector, Burlingame, CA). An avidin–alkaline phosphatase–fast red reagent or the peroxidase–3,3′-diaminobenzidine system (Vectastain ABC kit; Vector) was used to visualize the antibody stains. Normal IgG (Sigma) was used as the control for the immunostains.

Measurement of collagen content by Sircol assay.

Punch skin biopsy samples (6 mm²) were taken and their non-crosslinked fibrillar collagen was measured using the Sircol colorimetric assay (Biocolor, Belfast, UK). Skin samples were obtained from the lower backs of age- and sex-matched genotypically identical siblings injected with 4-OHT or sunflower oil. The skin biopsy samples were cut into small pieces and homogenized in 0.5M acetic acid with 1:20 (weight/weight) pepsin overnight at 4°C. The samples were centrifuged, and 100 μl of supernatant was used in the analysis with the Sircol dye reagent. Dye–collagen complexes were solubilized with alkali and measured in a spectrophotometer at 540 nm. To determine exact collagen content per biopsy sample, standard curves were used. Data were expressed as mean collagen content per biopsy sample normalized to DNA content.

Cell culture and transient transfections.

Fibroblasts growing out of skin explants described previously (17) at passage 2 or 3 were used for immunofluorescence, Northern blot, Western blot, and transfection analyses. Primary skin fibroblasts were transfected with 0.6 μg of reporter plasmid (PAI-1/luc or 3TP/luc [18,19]) and 0.4 μg of pSV-β-galactosidase (pSV-β-gal; Promega, Madison, WI) with FuGene (Roche, Nutley, NJ). After 24 hours, TGFβ was added to the medium at a concentration of 4 ng/ml. Luciferase activity was measured after an additional 24 hours. Cells were cotransfected with a vector expressing β-gal, and luciferase activity was normalized for β-gal activity measured with a β-gal enzyme assay system. The total amount of DNA was kept constant using empty vector DNA.

Northern blot analysis.

Total cellular RNA was isolated from skin fibroblasts (Qiagen, Chatsworth, CA). Total RNA transferred to nylon membranes was hybridized with 1 × 10⁶ counts per minute/ml of ³²P-labeled cDNA probes to connective tissue growth factor (CTGF), tissue inhibitor of metalloproteinases 1 (TIMP-1), TIMP-2 type I collagen, fibronectin, SPARC (secreted protein, acidic and rich in cysteine), osteopontin, Smad7, and decorin. The blots were washed and exposed to XAR-5 film (Eastman Kodak, Rochester, NY) at –70°C.

Western blot analysis.

Primary skin fibroblasts were lysed and fractionated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis. Proteins transferred to nitrocellulose membranes were incubated with p-Smad2/3 and Smad4 (both from Santa Cruz Biotechnology), p-p38 and p–ERK-1/2 (both from Cell Signaling Technology, Beverly, MA), and actin (Sigma). Blots were incubated with anti-mouse or anti-goat horseradish peroxidase–conjugated secondary antibody, developed using ECL reagents (NEN Life Science Products, Boston, MA), and exposed to Kodak XAR-5 film.

RESULTS

Controlled up-regulation of constitutively active TGFβRI specifically in fibroblasts.

To generate mice in which the expression of constitutively active TGFβRI could be induced in fibroblastic cells after birth (TBR1^CA-1 knockin mice), we inserted constitutively active TGFβRI in the ROSA26 locus by homologous recombination in mouse ES cells. In the targeting vector, the DNA for constitutively active TGFβRI was cloned 3′ of a transcription stop cassette, which itself is flanked by loxP sites. This vector was targeted to the ubiquitously active ROSA26 locus in mice (16) (Figure 1A). The constitutively active TGFβRI cDNA contains a hemagglutinin tag to allow detection of the knockin gene.

Figure 1. Targeting strategy for the insertion of constitutively active transforming growth factor β receptor type I (TBR1^CA) in the ROSA26 locus. A, Structure of the genomic R26R (ROSA26) locus, targeting vector, and targeted allele. Solid boxes indicate exons; open boxes indicate 3′-untranslated region and poly A sequence; blue lines depict intronic sequences. Brown and green lines indicate DNA fragments detected by Southern blot analysis. EI = Eco RI; EV = Eco RV; SA = splice acceptor; STOP = transcription stop cassette; TBR1^CAHA = hemagglutinin-tagged constitutively active transforming growth factor β receptor type I. B, Southern blot analysis of genomic DNA of embryonic stem (ES) cells to identify the clones with the knockin (KI) gene for ES cell injection. Left, Eco RI digestion; right, Eco RV digestion. WT = wild type. C, Schematic representation of the Col1a2 Cre-ER transgene and of the constitutive activation of transforming growth factor β receptor type I after injections of 4-hydroxytamoxifen (4-OHT). (T) = tamoxifen. D, Immunofluorescence of hemagglutinin-tagged constitutively active transforming growth factor β receptor type I transgene expression in skin fibroblasts using an antihemagglutinin antibody. Constitutively active transforming growth factor β receptor type I knockin gene expression was detected in fibroblasts from mice injected with 4-OHT, but not in control skin fibroblasts derived from mice injected with oil. Bar = 50 μm. E, Gross skin morphology of 5-month-old littermates injected with oil or 4-OHT.

We anticipated that no expression of constitutively active TGFβRI would occur in the absence of active Cre recombinase. Southern blot analysis of the genomic DNA of ES cells was performed, and genomic DNA was hybridized with the 3′ probe. ES cell clones with the knockin allele generated a WT fragment of 15 kb and a knockin fragment of 10 kb. Similarly, Southern blot analysis of genomic DNA digested with Eco RI and probed with a 5′ probe detected a WT allele and a mutant allele of 11 kb and 2 kb, respectively (Figure 1B). We had previously generated transgenic mice in which a fibroblast-specific enhancer of the mouse Col1a2 gene directs expression of the polypeptide consisting of a fusion between Cre recombinase and a mutant ligand-binding domain of the estrogen receptor (20). In these Col1a2 Cre-ER–transgenic mice, administration of 4-OHT activates Cre recombinase in fibroblastic cells. The TBR1^CA-knockin mice were crossed with the Col1a2 Cre-ER–transgenic mice, and offspring positive for constitutively active TGFβRI and the Cre-ER fusion protein (TBR1^CA; Cre-ER mice) were injected with 4-OHT for 5 consecutive days beginning 2 weeks after birth (Figure 1C). The duration and dose of 4-OHT injections were chosen based on earlier work by Zheng et al (21).

To determine whether expression of constitutively active TGFβRI in the TBR1^CA; Cre-ER mice occurred in fibroblasts only after injection of 4-OHT, immunofluorescence with antihemagglutinin antibody was performed on cultured skin fibroblasts. Expression of hemagglutinin was observed only in fibroblasts from 4-OHT–injected mice; no expression occurred in control fibroblasts from oil-injected mice with the identical genotype, suggesting that the knockin gene for constitutively active TGFβRI was specifically activated by 4-OHT injection (Figure 1D).

Initial gross examination of the mice did not reveal any morphologic change or abnormality. The animals were fertile and grew to their normal size, except that they were leaner than their controls. For mice treated with oil, their mean ± SD weight at 5 months was 33.6 ± 2.45 gm, whereas those injected with 4-OHT had a mean ± SD weight of 28 ± 2.0 gm. The difference in weight between the 2 groups was significant (P < 0.001). When the animals were killed at age 5 months, a complete serum chemistry and hematologic profile of TBR1^CA; Cre-ER mice showed no abnormality in the levels of albumin, gamma globulins, electrolytes, alanine aminotransferase and aspartate aminotransferase hepatic enzymes, blood urea nitrogen, and creatinine or in the numbers of white or red blood cells or in the levels of hematocrit and hemoglobin. Soft radiographs of whole animals did not reveal any bone anomalies in the TBR1^CA; Cre-ER mice. The mice developed hair loss on the back, on the head, around the ears and nose, along the sides, and on the chest. A representative mouse is shown in Figure 1E. The skin felt thicker and had a rougher texture than that of littermate controls.

Increased dermal fibrosis progressing with age.

Dermal fibrosis in TBR1^CA; Cre-ER mice injected with 4-OHT was compared with that in age- and sex-matched controls of identical genotype injected with oil. Sex effects were assessed, and same-sex comparison was performed. Skin biopsy samples from similar anatomic locations were examined in both groups of mice 1.5 months, 3 months, and 5 months after injection. As early as 1 month after injection with 4-OHT, mice started to show fibrosis of the dermis, which progressively increased at 3 months and at 5 months, whereas the dermis in control mice injected with oil was identical to that of WT mice. By 5 months, pronounced changes were seen in the dermal architecture of the skin, including thinning of the epidermal layer and subcutaneous fat and dramatic increases in the thickness of the dermal layer. The skin was fragile, and histology showed that the epidermal and dermal layers were often separated. Staining with Masson's trichrome indicated significant dermal accumulation of collagen in 4-OHT–injected TBR1^CA; Cre-ER mice (Figure 2). Image analysis of the intensities of Masson's trichrome staining in the dermis of skin sections using MetaMorph software (data not shown) indicated an ∼2-fold increase per square micrometer at each of the 3 time points shown in Figure 2. We did not observe any sex-based differences in the extent of the dermal fibrosis.

Figure 2. Generalized and pronounced dermal fibrosis resulting from sustained overexpression of constitutively active transforming growth factor β receptor type I in skin fibroblasts in vivo. Skin biopsy samples were taken from the lower backs of age- and sex-matched littermates that had been injected with oil (control) or with 4-hydroxytamoxifen (4-OHT) at age 1.5 months, 3 months, and 5 months. The skin sections were stained with hematoxylin and eosin (top 2 rows) and Masson's trichrome (bottom 2 rows). Progressive dermal thickening is evident at 1.5 months, 3 months, and 5 months after 4-OHT injections. Bar = 200 μm.

Quantitation of total collagen in 6-mm² skin biopsy samples showed a 2.3-fold increase in 4-OHT–injected TBR1^CA; Cre-ER mice compared with control mice (Figure 3A). Immunohistochemistry also showed increased accumulation of both type I and type III collagens in skin samples from 4-OHT–injected TBR1^CA; Cre-ER mice (Figures 3C and E) compared with control mice (Figures 3B and D), and increased type I collagen deposition in the interstitium of the kidney in 4-OHT–injected TBR1^CA; Cre-ER mice (Figure 3G) compared with control mice (Figure 3F).

Figure 3. Increased collagen synthesis in mice expressing constitutively active transforming growth factor β receptor type I and Col1a2 Cre-ER fusion protein (TBR1^CA; Cre-ER mice) injected with 4-hydroxytamoxifen (4-OHT). A, Non-crosslinked fibrillar collagen content of skin was measured using the Sircol assay. Increased collagen synthesis and accumulation were evident in skin biopsy samples from 4-OHT–injected TBR1^CA; Cre-ER mice compared with those from littermate controls injected with oil. Values are the mean and SD. B–E, Immunohistochemistry of skin biopsy samples using antibodies to either type I collagen (B and C) or type III collagen (D and E) shows increased collagen expression and accumulation of both types of collagen in areas of dermal fibrosis in mice injected with 4-OHT compared with littermate controls injected with oil. Bar = 200 μm. F and G, Immunohistochemistry of kidney sections with antibody to type I collagen shows increased interstitial collagen in mutant mice injected with 4-OHT (G) compared with control mice (F). Bar = 50 μm.

Ultrastructural analysis of collagen fibrils by transmission electron microscopy revealed considerable disorganization of collagen bundles and a great variability in shape and size of the collagen fibrils in the dermis of 4-OHT–injected mice (results not shown) compared with controls. Moreover, thin and large collagen fibrils intermingled in the papillary dermis and reticular dermis. In contrast, skin of control mice showed collagen fibrils with fairly uniform, circular cross sections (results not shown).

Vascular changes associated with increased TGFβ expression.

Deposition of collagen in the walls of small arteries causing irreversible damage is a hallmark of SSc (22). In 4-OHT–injected TBR1^CA; Cre-ER mice, the heart and aorta appeared to be normal (results not shown), but the walls of the smaller arteries in the lung were thickened (Figures 4C and D) compared with those of arteries from control mice (Figures 4A and B). From the trichrome staining of the lung blood vessels, there appeared to be hypertrophy of the smooth muscle layers, which is indicative of pulmonary arterial hypertension, a hallmark of SSc. The ratios of vessel wall thickness to vessel circumference and of vessel wall thickness to diameter showed a 1.5–2-fold increase (P < 0.0005 and P < 0.0004, respectively) (Figure 4E). The same was true for vessels of the kidney and adrenal glands (Figure 4F). Blinded observers obtained measurements of blood vessel wall thickness.

Figure 4. Vascular changes in small blood vessels associated with activated transforming growth factor β pathway in fibroblasts. A–D, Comparison of small blood vessel wall width in the lungs from TBR1^CA; Cre-ER mice injected with 4-OHT (C and D) or with oil (A and B). The lung sections were stained with Masson's trichrome (original magnification × 100). E, Measurement of vessel walls was done using MetaMorph software. The ratio of vessel wall width to circumference is shown at the left (∗ = P < 0.0005), and the ratio of vessel wall width to diameter is shown at the right (∗ = P < 0.0004). Values are the mean and SD. F, A similar increase in small blood vessel wall thickness was observed in the kidney and adrenal glands of TBR1^CA; Cre-ER mice injected with 4-OHT compared with those injected with oil. Kidney sections were stained with hematoxylin and eosin. Bar = 50 μm. G, Increased staining with antibody against von Willebrand factor, a marker for endothelial cells, is associated with an increase in blood vessel wall thickness in lung sections. Bar = 125 μm. See Figure 3 for definitions.

Patients with SSc show endothelial cell activation and damage, with increased plasma levels of vWF and E-selectin (23–25). Immunohistochemistry of lung blood vessels in 4-OHT–treated TBR1^CA; Cre-ER mice showed increased vWF expression (Figure 4G), suggesting that sustained activation of TGFβ signaling in fibroblasts could initiate endothelial cell damage in these mice.

Increased expression of downstream TGFβ targets.

To explore the mechanism associated with dermal fibrosis, expression of TGFβ downstream targets in fibroblasts from skin explants was examined by Northern blot analysis (Figure 5A). One of the immediate TGFβ targets, CTGF, also a key player in fibrosis (26, 27), was highly overexpressed in fibroblasts from mice injected with 4-OHT, whereas there was very little expression in control cells. Other TGFβ targets, such as TIMP-1, also showed strongly increased expression. RNA levels of type I collagen, fibronectin, SPARC, osteopontin, and Smad7 were all higher in skin fibroblasts from 4-OHT–treated mice than in those from controls, suggesting that TGFβ downstream targets were activated and contributed to the skin phenotype. Interestingly, decorin, a small leucine-rich protein important for collagen fibrillogenesis (28), was dramatically down-regulated.

Figure 5. Downstream signaling targets activated by constitutive transforming growth factor β receptor type I (TGFβRI). A, Northern blot analysis to measure expression of target genes of the activated TGFβ pathway in primary dermal fibroblasts of TBR1^CA; Cre-ER mice injected with oil or 4-OHT. B, Immunofluorescence of primary dermal fibroblasts from 4-OHT–injected TBR1^CA; Cre-ER mice shows nuclear localization of p-Smad2/3 (red; top) and cytoplasmic localization of Smad7 (green; bottom), whereas littermate control mice with the same genotype injected with oil exhibited no p-Smad2/3 signaling in the nucleus or Smad7 expression in the cytoplasm. Bar = 50 μm. C, Western blot of skin fibroblasts from TBR1^CA; Cre-ER mice injected with 4-OHT or oil using antibody against Smad4. D, Western blot of primary skin fibroblasts from 4-OHT–injected mice shows increased phosphorylation of ERK-1/2 and p38. Antibody against actin was used as a loading control for the Western blots in C and D (results not shown). No differences were observed with an antibody against actin. E, Luciferase activity of the TGFβ target promoters in primary dermal fibroblasts. Primary fibroblasts from the skin of control mice and mice injected with 4-OHT were used. The PAI-1/Luc (PAI-1/L) or 3TP/Luc (3TP/L) reporter plasmids were transfected together with a vector directing β-galactosidase expression as an internal control. Luciferase activity was measured using a luciferase assay system that was normalized to β-galactosidase activity. Values are the mean and SD from a minimum of 3 experiments. See Figure 3 for other definitions.

Activation of the TGFβ pathway also caused nuclear translocation of p-Smad2/3 in skin fibroblasts of TBR1^CA; Cre-ER mice injected with 4-OHT (Figure 5B, top right), whereas p-Smad2/3 was undetectable in control cells (Figure 5B, top left). In contrast, the levels of Smad4 were similar in both types of fibroblasts (Figure 5C). Thus, increased receptor-activated Smad phosphorylation and nuclear import provide evidence for activation of Smad signaling. We also observed increased accumulation of Smad7 (Figure 5B, bottom right), a critical negative regulator of the TGFβ pathway (29) normally induced by TGFβ in fibroblasts (30). The degree of activation of the TGFβ pathway in these mice is probably the result of the relative concentrations of phosphorylated Smad2/3 and Smad7.

In addition to activation of the Smad pathway, we observed phosphorylation of ERK-1/2 and of p38 in fibroblasts from 4-OHT–injected TBR1^CA; Cre-ER mice, but not in control mice. Thus, constitutive activation of TGFβRI in fibroblasts increased the activity of the MAP kinase (MAPK) and p38 pathways (Figure 5D).

In addition to downstream targets of TGFβ, we examined the activity of two TGFβ-regulated target promoters, plasminogen activator inhibitor 1 (PAI-1) and 3TP. There is a low basal activity of PAI-1 and 3TP luciferase reporters in control dermal fibroblasts that is increased 3-fold and 4-fold, respectively, upon addition of exogenous TGFβ1. In the dermal fibroblasts from the 4-OHT–injected mice, basal expression was markedly elevated, and addition of exogenous TGFβ1 to these cells increased expression of promoter reporter constructs less than 1.5-fold (Figure 5E). This suggests that in these cells, the TGFβ pathway was already activated and that exogenous TGFβ did not much increase this activity.

In SSc the dermis shows a heterogeneous population of fibroblasts, some of which have differentiated into myofibroblasts. The skin fibroblasts from control mice showed typical spindle-shaped morphology and were negative for α-SMA. In contrast, ∼15–20% of skin fibroblasts from 4-OHT–injected mice stained positively for α-SMA and had a flattened appearance, suggesting transformation into myofibroblasts (Figure 6). Similar to dermal fibroblasts from SSc patients, the dermal fibroblasts from TBR1^CA; Cre-ER mice consisted of a heterogeneous population of normal fibroblasts and myofibroblasts.

Figure 6. Transformation of dermal fibroblasts. A heterogeneous population of fibroblasts was observed in primary dermal fibroblasts from TBR1^CA; Cre-ER mice injected with 4-OHT. Immunofluorescence with α-smooth muscle actin showed a population of cells that had transformed into myofibroblasts in these mice, whereas cells from oil-injected TBR1^CA; Cre-ER mice lacked myofibroblasts. Nuclei were stained with 4′,6-diamidino-2-phenylindole. Bar = 50 μm. See Figure 3 for definitions.

DISCUSSION

In this study, we have demonstrated that the TBRI^CA; Cre-ER mice recapitulate key pathologic features of SSc when they are injected with 4-OHT, which activated TGFβ signaling. These features included a marked skin fibrosis that increased with age and a thickening of the walls of small arteries in lung and kidney. In vivo and in vitro measurements of collagen indicated a marked increase in collagen accumulation and collagen gene expression. Our results are consistent with a central role of TGFβ in disease pathogenesis (8, 9), and they identify the fibroblast as a critical target cell (11–13). A logical implication of this view is that other pathogenic processes such as immunologic activation or endothelial cell dysfunction, both of which are observed early in scleroderma, lead to TGFβ-mediated activation of fibroblasts in vivo. The conditional activation of signaling using the Cre/loxP system with deletion of a transcription stop cassette is a powerful method, since it allows high-level activation that is transmissible to daughter cells at mitosis and does not depend upon the continued presence of the tamoxifen ligand.

Explanted dermal fibroblasts showed many of the hallmark biochemical properties of SSc. Key features included myofibroblast differentiation, ECM overproduction, and constitutive overexpression of a number of downstream TGFβ targets. Two TGFβ-responsive Smad-dependent promoters were also up-regulated in the mutant fibroblasts. We therefore hypothesize that the activation of the genes observed in the mutant fibroblasts is mediated by the constitutive activity of the Smad pathway. As in fibroblasts from SSc patients (31), activation of the MAPK and p38 pathways in the fibroblasts of mutant mice was also observed.

The mutant mice did not show fibrotic lesions in the lungs or renal insufficiency, which are often late manifestations of human scleroderma. One possible explanation is that by bypassing an initial inflammatory component in our mouse model, we have eliminated a critical initiating event that leads to the lung pathology seen in scleroderma. Additionally, our mice are housed in a specific pathogen–free facility and are completely free of any pathogenic agents, while in the real world, exposure to such agents might be a possible injury or trigger in initiating the cascade of events that culminate in the inflammatory/autoimmune features of scleroderma. We have, however, recently observed that intratracheal administration of bleomycin in the same 4-OHT–treated mutant mice causes a more pronounced fibrotic response in the lungs than in control mice (Sonnylal S: unpublished observations). It is possible that additional injuries are needed to trigger renal insufficiency in TBR1^CA; Cre-ER mice.

In fibroblasts derived from mutant mice, there was a marked increase in the levels of CTGF RNA. Similar increases have been seen in fibroblasts from scleroderma patients and after TGFβ stimulation of normal fibroblasts (26). Like TGFβ, CTGF can induce matrix production in fibroblasts (32, 33). We have recently generated transgenic mice that overexpress CTGF in fibroblastic cells. These mice also show generalized skin fibrosis (Sonnylal S: unpublished observations). Comparison of the changes in specific RNA levels and of the activities of specific signaling pathways in the fibroblasts of these mice with those in which constitutively active TGFβRI is expressed in fibroblasts should aid in understanding the mechanisms of CTGF function.

During wound healing, TGFβ and other cytokines are known to promote ECM production. Any disruption or excessive activity of these signals can result either in impairment or in excess tissue formation, resulting in fibrosis. In tissue fibrosis, the fine balance that exists between the expression of matrix metalloproteinases and the expression of TIMPs is impaired (34). Studies investigating the association of TIMPs with fibrosis in scleroderma have found elevation of TIMP-1 levels increasing with the severity of the disease (35). The increased levels of TIMP-1 expression in fibroblasts that express constitutively active TGFβRI are likely to have contributed substantially to the fibrotic phenotype of our mutant mice.

Animal models have been very valuable in defining pathogenic mechanisms as well as testing therapeutic interventions in human disease. Although there are several available models of SSc, each of these incompletely represents the human condition (17, 36–40). For example, the tight skin mouse develops dermal fibrosis, but the pathogenic mechanisms are not defined (36). In addition, recent reexamination of these mice has challenged their usefulness as a model of scleroderma (41). Subcutaneous injection of bleomycin induces transient skin sclerosis only in animals with a susceptible genetic background, and this diminishes once the injurious stimulus is removed (39). Minimal mismatch graft-versus-host disease has been used to investigate the potential for TGFβ-blocking therapies but has proven variable in severity and is generally self-limiting (40).

The mice with the constitutively active fibroblast-specific TGFβRI develop 2 cardinal features of established SSc, dermal fibrosis and structural vasculopathy. In addition, there is evidence of generalized abnormalities in ECM deposition extending beyond sites that are clinically involved, similar to what is now well recognized in SSc. This mouse model offers substantial advantages that include a defined nature, the primary defect being activation of TGFβ signaling in fibroblasts. Our mouse model can be regulated—this allows the developmental consequences during embryogenesis of altered TGFβ signaling to be avoided. While we appreciate that the potentially critical early immunologic and inflammatory components of human SSc are not reproduced in this model, our findings provide important direct evidence that activating TGFβ signaling in just the fibroblastic lineage replicates key fibrotic features of SSc. This suggests that in vivo in human disease, one of the eventual consequences of these (vascular, immunologic) events is activation of TGFβ signaling in fibroblasts.

By circumventing these events, our model is uniquely placed to study the late-stage fibrotic phase of SSc, which is particularly difficult to treat. It allows long-term sustained pathway activation, and so, overcomes some of the limitations of induced models. This model also offers substantial potential for the evaluation of therapies that target TGFβ activation or downstream consequences of this activation, including the effects of secondary mediators such as CTGF and others. The ability to induce the disease phenotype at defined time points will allow prevention as well as treatment strategies to be examined. These mice will allow other injurious stimuli to be evaluated, including vascular, immunologic, or pulmonary epithelial injury. For example, epithelial abnormalities may be important, and these may have different effects during development and during the postnatal period. These mice show much-increased susceptibility to bleomycin-induced lung disease (results not shown), providing strong validation of our model in SSc.

In conclusion, the characterization of a novel mouse model, in which TGFβ signaling can be activated selectively in fibroblasts at defined postnatal time points, provides strong direct evidence of a pivotal role of this growth factor in fibrotic diseases and scleroderma. Future modifications of this model are likely to result in an even more comprehensive clinical scleroderma phenotype. Our model, which provides insight into pathogenic mechanisms in this disease, will also be used as a platform to evaluate therapies targeting TGFβ activation.

AUTHOR CONTRIBUTIONS

Dr. de Crombrugghe had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Drs. Sonnylal, Denton, Zheng, Behringer, and de Crombrugghe.

Acquisition of data. Drs. Sonnylal, Denton, Zheng, Keene, and He, Mr. Adams, and Ms Deng.

Analysis and interpretation of data. Drs. Sonnylal, Denton, Keene, VanPelt, Geng, and de Crombrugghe.

Manuscript preparation. Drs. Sonnylal, Denton, and de Crombrugghe.

Statistical analysis. Dr. Sonnylal.

Acknowledgements

We thank Dr. Phillip Soriano for the ROSA26 targeting vector, Dr. Joan Massague for the constitutively active TGFβRI cDNA, Zhaoping Zhang for injecting ES cells into blastocysts, Nehal Porecha and Ying Wang for technical help, and Dr. Frank Arnett for valuable discussions.