Hedgehog (Hh) and transforming growth factor-β (TGF-β) family members are involved in numerous overlapping processes during embryonic development, hair cycle, and cancer. Herein, we show that TGF-β induces the expression of the Hh signaling molecules Gli1 and Gli2 in various human cell types, including normal fibroblasts and keratinocytes, as well as various cancer cell lines. Gli2 induction by TGF-β is rapid, independent from Hh receptor signaling, and requires a functional Smad pathway. Gli1 expression is subsequently activated in a Gli2-dependent manner. In transgenic mice overexpressing TGF-β1 in the skin, Gli1 and Gli2 expression is also elevated and depends on Smad3. In pancreatic adenocarcinoma cell lines resistant to Hh inhibition, pharmacologic blockade of TGF-β signaling leads to repression of cell proliferation accompanied with a reduction in Gli2 expression. We thus identify TGF-β as a potent transcriptional inducer of Gli transcription factors. Targeting the cooperation of Hh and TGF-β signaling may provide new therapeutic opportunities for cancer treatment. [Cancer Res 2007;67(14):6981–6]
- signal transduction
The hedgehog (Hh) signaling pathway is critical for stem cell maintenance, embryonic patterning, and growth in both invertebrates and vertebrates ( 1). Deregulation of the Hh pathway is a characteristic trait of several pathologic states, including developmental syndromes with high proneness to cancer ( 1, 2). Cellular responses to the Hh signal are controlled by two transmembrane proteins, the tumor suppressor Patched-1 (Ptch) and the oncoprotein Smoothened (Smo; refs. 2, 3). The latter has homology to G protein–coupled receptors and transduces the Hh signal. In the absence of Hh, Ptch maintains Smo in an inactive state, thus silencing intracellular signaling. With the binding of Hh, Ptch inhibition of Smo is released and the signal is transduced. The transcriptional response to Hh signaling is mediated by a family of zinc-finger transcription factors that comprises the Ci protein in Drosophila and the three closely related Gli proteins in vertebrates, Gli1, Gli2, and Gli3 ( 4). Gli2 is thought to function upstream of Gli1 and to be the primary mediator of Hh signaling ( 5), inducing Gli1 expression via direct binding to its promoter region ( 6). Gli transcription factors regulate multiple cellular functions associated with malignant transformation, such as cell cycle progression and apoptosis ( 4, 7).
The importance of the Hh signaling pathway in tumorigenesis was established through the discovery of inactivating mutations in the Ptch gene in patients with familial (Gorlin's syndrome) basal cell carcinomas (BCC) and sporadic BCC ( 8, 9). Other tumors exhibit inappropriate Hh pathway activation. For example, some esophageal squamous cell sarcomas and transitional cell carcinomas of the bladder may carry loss-of-function mutations of the Ptch gene ( 10), whereas gain-of-function mutations of Smo have been identified in a subset of small cell lung carcinoma ( 11). Additionally, overexpression of the main Hh member Sonic Hh (Shh), leading to activation of Smo, has been identified in some gastrointestinal cancers ( 12) and pancreatic adenocarcinomas ( 13).
Similar to Hh members, transforming growth factor-β (TGF-β) has emerged as a family of growth factors involved in various essential physiologic processes that include embryonic development, tissue repair, cell growth control, and differentiation. TGF-β isoforms are expressed in a variety of tumor types and contribute to the aggressiveness and progression of neoplasms ( 14, 15). TGF-β members signal via membrane-bound heteromeric serine-threonine kinase receptor complexes. In most cell types, TGF-β binds to TβRII in combination with TβRI, also known as ALK5 ( 16). Receptor activation by TGF-β leads to phosphorylation of cytoplasmic proteins of the Smad family. Receptor-associated Smads, Smad2 and Smad3, then heteromerize with Smad4, translocate into the nucleus, and act as transcription factors to regulate target gene expression. Smad3 is thought to contribute most Smad-dependent responses to TGF-β in the adult, whereas Smad2 is critical during embryogenesis ( 17, 18).
In this study, we have examined the capacity of TGF-β to modulate the expression of the Hh signaling molecules Gli1 and Gli2. We provide definitive evidence, both in vitro and in vivo, for Smad-dependent activation of Gli2 expression and, consequently, that of Gli1, thereby identifying TGF-β as a cytokine ubiquitously capable of activating, enhancing, or prolonging Hh signals. In addition, we show that some cyclopamine-resistant pancreatic adenocarcinoma cell lines are growth inhibited by a small-molecule inhibitor of TGF-β signaling.
Materials and Methods
Cell cultures and reagents. Primary human dermal fibroblasts, WI-26 human transformed lung fibroblasts, HaCaT immortalized human keratinocytes, MDA-MB-231 and MDA-MB-468 breast carcinoma cell lines, and PANC-1 human pancreatic adenocarcinoma cells were all maintained in DMEM with 10% fetal bovine serum and antibiotics (Invitrogen). When indicated, cells were serum starved for 16 h and treated with human recombinant TGF-β1 (5 ng/mL; referred to as TGF-β) and/or human recombinant NH2-terminal Shh peptide (1.5 μg/mL), both purchased from R&D Systems. Cyclopamine, cycloheximide, and the kinase inhibitors SB431542, SB203580, PD98059, and SP600125 were obtained from Euromedex. LY294002 was from Calbiochem-Merck. Small interfering RNAs (siRNA) were purchased from Ambion/Applied Biosystems and transfected into cells using the RNAiFect reagent (Qiagen). Transfection of MDA-MB-468 cells with Smad3 ( 19) and Smad4 ( 20) expression vectors was done using an electroporation kit (Amaxa Biosystems). Cell growth was estimated by 3-(4-5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium salt (MTS) assay, using a specific kit (Promega) according to the manufacturer's protocol.
Multiplex PCR and real-time PCR. Total RNA was prepared using an RNeasy mini kit (Qiagen). Reverse transcription (Invitrogen) was done on genomic DNA-free RNA using oligo(dT) as primer. cDNAs are then used in multiplex PCR according to Qiagen recommendations. Real-time PCR was done with either a Power SYBR Green Mix or Taqman probes (mouse experiments) on an AB7300 apparatus (Applied Biosystems). The absolute copy number for each mRNA was normalized to the absolute cyclophilin A mRNA copy number. PCR primer sequences and conditions are available on request.
Western blot analyses. Cells were lysed with ice-cold lysis buffer [20 mmol/L HEPES (pH 7.9), 420 mmol/L NaCl, 0.5% NP40, 25% glycerol, 1.5 mmol/L MgCl2, 0.2 mmol/L EDTA], rapidly frozen and thawed, rotated 30 min at 4°C, and centrifuged for 30 min. Total protein (100 μg) was separated by SDS-PAGE and blotted onto nitrocellulose membranes. After saturation with TBS+0.1% Tween 20+5% dry milk, membranes were incubated with either anti-Gli2 (Santa Cruz Biotechnology, Inc.) or anti-actin (Zymed) antibodies. Detection was done using horseradish peroxidase–conjugated secondary antibodies (Santa Cruz Biotechnology) and revealed with enhanced chemiluminescence (Amersham Biosciences).
Transgenic mice. To determine the capacity of TGF-β to induce Gli expression in vivo, we used a transgenic mouse model constitutively expressing TGF-β1 in the epidermis under the control of the keratin 5 promoter (K5-TGF-β1; ref. 21), as well as an inducible gene-switch mouse model allowing inducible expression of TGF-β1 in the epidermis under the control of the progesterone receptor ( 22). To determine the relative implication of Smad2 and Smad3 in mediating TGF-β effects on Gli expression, K5-TGF-β1 mice were crossed with either Smad2+/− or Smad3+/− mice ( 23) and Gli expression was measured in compound heterozygote animals.
Results and Discussion
TGF-β activates Gli1 and Gli2 expression in various cell types. We first examined whether TGF-β has a direct effect on the expression of various components of the Hh signaling cascade. For this purpose, cultures of human neonatal dermal fibroblasts (NHDF; Fig. 1A ), HaCaT keratinocytes ( Fig. 1B), and MDA-MB-231 breast carcinoma cells ( Fig. 1C) were incubated with TGF-β and RNA was extracted at various time points. Remarkable conservation of the modulation of Gli1 and Gli2 expression was observed in all cell types tested. Specifically, rapid and persistent induction of Gli2 was observed in response to TGF-β, peaking at 2 to 8 h and remaining at levels significantly higher than their basal expression state up to 16 to 24 h. On the other hand, delayed induction of Gli1 was observed, with a maximum ∼48 h after TGF-β addition. Similar patterns of Gli1 and Gli2 regulation by TGF-β were also identified in human adult primary skin fibroblasts, immortalized lung fibroblasts (WI-26 cell line) and keratinocytes (NCTC2544 cell line), as well as other human cancer cell lines, including pancreatic adenocarcinoma, glioblastoma, and melanoma (data not shown).
Western blot analysis of total Gli2 production in dermal fibroblasts showed undetectable expression levels in unstimulated cultures. Gli2 protein became detectable 6 h after addition of TGF-β and accumulated over 48 h of incubation ( Fig. 1D). Given the high levels of Gli2 protein detected at the 48-h time point, it is possible that TGF-β may stabilize GLI2 protein, in addition to increasing gene expression.
Of note, Gli3 (see multiplex PCR panels), Ptch, and Smo (data not shown) expression in response to TGF-β showed minimal variation.
Activation of Gli expression by TGF-β does not involve the Ptch/Smo axis. To determine whether TGF-β–induced Gli1 and Gli2 expression was dependent on the Ptch/Smo axis, the effect of exogenous Shh and TGF-β on Gli expression by human dermal fibroblasts was tested in the absence or presence of the Smo inhibitor cyclopamine, known to prevent Gli activation by Shh ( 24). As shown in Fig. 2A , both TGF-β and Shh strongly elevated Gli1 mRNA steady-state levels. As expected, cyclopamine abrogated Shh-induced Gli expression ( Fig. 2A, lane 4 versus lane 3) but had no effect on TGF-β–mediated Gli1 activation ( Fig. 2A, lane 6 versus lane 5). Furthermore, when added together to the culture medium, Shh and TGF-β exerted additive effects on both Gli2 and Gli1 expression ( Fig. 2B). Under the same conditions, Ptch-1, a classic Shh target, was strongly induced by Shh, not by TGF-β ( Fig. 2B, top), suggesting that Ptch-1 modulation is not solely dependent on Gli expression. Interestingly, three distinct skin fibroblast strains derived from patients with Gorlin's syndrome that carry heterozygote somatic loss-of-function mutations of the Ptch-1 gene ( 25) responded to TGF-β with significant up-regulation of both Gli2 and Gli1 expression, with kinetics similar to those observed in non–Gorlin-related cells. Specifically, all three Gorlin fibroblast strains exhibited a 4- to 11-fold transient elevation of Gli2 mRNA 4 h after TGF-β stimulation ( Fig. 2C, left) followed by a 4- to 9-fold elevation of Gli1 mRNA steady-state levels 24 h after addition of TGF-β ( Fig. 2C, right).
Together, these results indicate that Shh and TGF-β are capable of inducing Gli expression via distinct mechanisms and that the mechanisms underlying TGF-β effects do not involve the Ptch/Smo axis.
Gli2 induction by TGF-β is a Smad3-dependent mechanism and mediates subsequent Gli1 activation. As a first attempt to elucidate the mechanisms by which TGF-β activates Gli1 and Gli2 expression, the possible implication of de novo protein synthesis was determined. As shown in Fig. 3A , the protein synthesis inhibitor cycloheximide abrogated Gli1 induction by TGF-β but had no inhibitory effect on Gli2 activation. Thus, the rapid elevation of Gli2 expression in response to TGF-β involves existing transactivators, whereas the delayed activation of Gli1 requires de novo synthesis of mediators.
Among the immediate transduction pathways activated by TGF-β, the most important and ubiquitous one is the Smad cascade ( 18), although evidence exists for activation of mitogen-activated protein kinases (MAPK), stress-activated protein kinase/c-Jun NH2-terminal kinase, extracellular signal-regulated kinase, and p38/MAPK and of the phosphatidylinositol 3-kinase (PI3K)/AKT pathways, downstream of the TGF-β receptors ( 26, 27). As a first approach to discriminate between all pathways, we tested the effects of pharmacologic inhibitors that specifically target the MAPK and PI3K pathways in parallel to a TβRI/ALK5 inhibitor on TGF-β–induced Gli2 expression. As shown in Fig. 3B, neither the MAPK inhibitors SB203580, PD98059, and SP600125 nor the PI3K inhibitor LY294002, which all effectively blocked their respective target pathway in these experiments (data not shown), affected TGF-β effect on Gli2 expression, whereas the TβRI/ALK5 inhibitor SB431542 fully abrogated the induction.
Next, we used the breast carcinoma cell line MDA-MB-468, which carries a large homozygous deletion within chromosome 18 encompassing the entire Smad4 gene ( 28) and exhibits deficient Smad-dependent transcription in response to TGF-β ( 19, 20). Gli2 expression in mock-transfected cells was barely detectable and was not elevated in response to TGF-β ( Fig. 3C, lane 2 versus lane 1). However, with Smad3/Smad4 complementation, Gli2 induction by TGF-β was fully restored in these cells ( Fig. 3C, lane 4 versus lane 3), thus unequivocally implicating the Smad pathway in Gli2 activation by TGF-β.
Further evidence for the need for intact Smad signaling came from Smad3 knockdown experiments: transfection of specific Smad3 siRNAs into WI-26 fibroblasts before TGF-β stimulation markedly impaired the induction of Gli2 gene expression by TGF-β, whereas control siRNAs did not ( Fig. 3D). Similar observations were made in HaCaT keratinocytes (data not shown).
Genetic approaches have shown that Gli2 plays the preeminent role in the transcriptional response to Hh signaling and is required for induction of Gli1 ( 5, 29). Because Gli2 induction by TGF-β precedes that of Gli1, we hypothesized that, similar to Hh signaling, late Gli1 induction by TGF-β may be dependent on early Gli2 activation. Consistent with this hypothesis, Gli2 knockdown with Gli2 siRNAs, which efficiently abrogated Gli2 induction by TGF-β ( Fig. 3E, right), prevented Gli1 induction in WI-26 human lung fibroblasts ( Fig. 3E, left). Similarly, overexpression of a dominant-negative mutant form of Gli2, mGli2-EN, which lacks transcriptional activity ( 30), also partially abrogated the induction of Gli1 by TGF-β (data not shown).
Together with the fact that cyclopamine did not prevent Gli1 induction by TGF-β (see Fig. 2A), these results show a Shh/Ptch/Smo–independent, Smad3-dependent, activation of Gli2 in response to TGF-β, which leads to delayed induction of Gli1. Thus, TGF-β is capable of conveying signals that were initially thought to be almost exclusively dependent on Hh factors.
TGF-β activates Gli1 and Gli2 expression in vivo in a Smad3-dependent manner. To determine whether TGF-β was capable of modulating Gli expression in vivo, two distinct transgenic mouse models overexpressing TGF-β1 in the epidermis (see Materials and Methods) were examined. Quantitative reverse transcription-PCR (RT-PCR) indicated that Gli1 expression was 5- to 6-fold higher in the epidermis from both K5-TGF-β1 mice with sustained TGF-β1 transgene expression ( 21) and gene-switch-TGF-β1 mice with acute TGF-β1 transgene induction ( 22) compared with the epidermis of control animals, whereas Gli2 expression was 12- to 15-fold higher than in controls ( Fig. 4A ). Next, to determine whether Smad signaling was implicated in Gli induction by TGF-β in vivo, Gli expression levels were measured in the epidermis of K5-TGF-β1 transgenic mice and of either K5-TGF-β1 × Smad2+/− or K5-TGF-β1 × Smad3+/− compound heterozygote mice. As shown in Fig. 4B, Smad3, not Smad2, heterozygosity suppressed high Gli expression in the skin of K5-TGF-β1 mice, consistent with the broad reduction in TGF-β target gene expression levels described previously ( 31). These data unequivocally show the implication of Smad3 in TGF-β–induced Gli2 and Gli1 expression in vivo, in accordance with the in vitro results described in this report that identify Smad3/Smad4–dependent Gli activation by TGF-β in cell lines of both mesenchymal and epithelial origins and consistent with the described role of Smad3 versus Smad2 in mediating most transcriptional responses to TGF-β ( 17, 18).
Pharmacologic inhibition of TGF-β signaling inhibits the growth of cyclopamine-resistant pancreatic adenocarcinoma cell lines. There is ample evidence that Gli genes possess activities that are independent from Hh signaling. For example, Gli2 and Gli3 are widely expressed in the developing embryo, including in regions that are far from Shh production ( 32), and may be expressed downstream of fibroblast growth factor signaling ( 33). In addition, mice expressing mutant forms of Gli2 or Gli3 exhibit defects that are likely not solely dependent on loss of Hh signaling ( 34). Thus, what is crucial for a given cell is the overall state of Gli expression and function, and Hh might just be one of several ways to regulate it ( 2). In this context, a growing number of reports indicates the potential therapeutic benefit of targeting either the TGF-β/Smad or Shh/Gli signaling pathways to counter the neoplastic process, consistent with their respective pro-oncogenic capacities. For example, cyclopamine and small-molecule Shh antagonists have shown efficacy in reducing tumor burden and inducing cancer cell apoptosis ( 13, 35, 36). Likewise, interfering with TGF-β signaling has shown promising results in preventing tumor development in a variety of tumor types ( 37– 42). To identify specific molecular signatures in a given tumor is critical to help determine the adequate therapeutic strategy ( 43). Interestingly, in a study involving numerous pancreatic carcinoma cell lines, inhibition of the Hh pathway by cyclopamine led to reduced proliferation and increased apoptosis only in a subset of cells. On the other hand, the authors identified another subset of cancer cells that was entirely resistant to cyclopamine, suggesting that their constitutive Hh activation, estimated as Gli1 or Ptch-1 expression, may not occur via Ptch/Smo ( 13). We thus tested the possibility that TGF-β/Smad signaling may contribute to constitutive Gli expression in the cyclopamine-resistant PANC-1 cell line ( 13) and whether pharmacologic inhibition of the TGF-β pathway may inhibit their proliferation. As shown in Fig. 5A , PANC-1 cells showed significantly impaired growth when treated with the TβRI/ALK5 inhibitor SB431542, not with cyclopamine, accompanied with decreased basal Gli2 expression, as measured by quantitative RT-PCR ( Fig. 5B), suggesting that autocrine TGF-β signaling may contribute to basal Gli2 expression. Similar results were obtained with the BxPC-3 cell line (data not shown). Transfection of PANC-1 cells with Gli2 siRNA efficiently reduced basal Gli1 expression ( Fig. 5C), PANC-1 cell growth ( Fig. 5D), and strongly attenuated the antiproliferative effect of SB431542 ( Fig. 5D). Together, these data indicate that the growth-inhibitory activity of SB431542 on PANC-1 cells is, at least in part, driven by a reduction in Gli expression.
Conclusion. We have identified TGF-β as a potent inducer of both Gli1 and Gli2 expression, independent from the Ptch/Smo axis. Most importantly, Gli2 induction is rapid, does not require de novo protein synthesis, and is a direct target of the Smad pathway. These results show unambiguously that TGF-β is capable of directly inducing signals that, until now, were thought to be exclusive mediators of Hh signaling and open new venues for efficient therapeutic approaches against cancer progression.
Grant support: Emile and Henriette Goutière donation, Institut National de la Sante et de la Recherche Medicale, Cancéropole Ile-de-France-INCa, France (A. Mauviel), NIH (X-J. Wang), Centre National de la Recherche Scientifique (T. Magnaldo), and the EC 6th framework Specific Targeted Research Project Tumor Host genomics and Dutch Cancer Society (P. ten Dijke). Association pour la Recherche sur le Cancer (France) postdoctoral fellowship (S. Dennler).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Dr. M. Hebrok (University of California, San Francisco, San Francisco, CA) for the generous gift of pancreatic adenocarcinoma cell lines.
- Received February 6, 2007.
- Revision received April 11, 2007.
- Accepted May 4, 2007.
- ©2007 American Association for Cancer Research.