Transforming growth factor β1 (TGF-β1) acts as a tumor suppressor at early stages of carcinogenesis, however, it has also been suggested to promote tumor progression at late stages. To determine at which stage and by what mechanisms this functional switch occurs, we have generated gene-switch-TGF-β1 mice in which TGF-β1 transgene expression can be induced in skin tumors at specific stages. These mice were exposed to a chemical carcinogenesis protocol, which allows tumorigenesis to develop in progressive stages from benign papillomas to malignant carcinomas. Remarkably, TGF-β1 transgene induction in papillomas rapidly induced metastasis. This function is in sharp contrast to its tumor suppressive effect when TGF-β1 transgene expression was induced early in the protocol. Transgenic papillomas exhibited down-regulation of TGF-β receptors and their signal transducer, the Smads, and loss of the invasion suppressor E-cadherin/catenin complex in the cell membrane. These molecules were lost only in malignant carcinomas in control mice at a much later stage. Furthermore, transgenic papillomas exhibited elevated expression of matrix metalloproteinases and increased angiogenesis. Our study suggests that TGF-β1 overexpression may directly induce tumor metastasis by initiating events necessary for invasion. Down-regulation of TGF-β signaling components in tumor epithelia selectively abolishes growth inhibition, thus, switching the role of TGF-β1 to a metastasis promoter.

The TGF4-β superfamily consists of multifunctional cytokines that regulate cell growth and differentiation, tissue remodeling, immune response, and angiogenesis (reviewed in Refs. 1, 2, 3, 4, 5, 6). The TGF-β direct family consists of three major isoforms in mammals: TGF-β1, -2, and 3. In most cell types, all three of the forms share similar biological activities. They act through two types of transmembrane serine/threonine kinase receptors, the type I and II TGF-β receptors (TGF-βRI and TGF-βRII). Upon ligand binding, TGF-βRI and TGF-βRII form tetraheteromers, which phosphorylate the downstream mediators Smad2 and/or Smad3. Smad2 and/or Smad3 then form heteromeric complexes with Smad4 and translocate to the nucleus to act as transcription factors (reviewed in Ref. 7). Smad6 or Smad7 can inhibit the TGF-β signaling cascade either by blocking phosphorylation and subsequent nuclear translocation of receptor-specific Smads (reviewed in Ref. 7) or by an increase in TGF-βR degradation via specific ubiquitin-proteasome pathways (8).

The TGF-β signaling pathway is one of the most important mechanisms for tumor suppression. Mice hemizygous for the TGF-β1 gene showed accelerated chemical carcinogenesis in the liver and lung indicating that TGF-β1 is a tumor suppressor with haploid insufficiency (9). The relevance of the TGF-β signaling pathway to the development of primary epithelial tumors in humans has been additionally substantiated by the discovery of mutations in TGF-βR and Smads (reviewed in Ref. 7). Paradoxically, many malignant human tumors express high levels of TGF-β, suggesting a potential tumor promoting role for TGF-β in late-stage carcinogenesis (reviewed in Refs. 10, 11). This is additionally supported by experiments on animal models. For instance, expression of the TGF-β1 transgene in the liver accelerated spontaneous and chemically induced hepatocarcinogenesis (12).

The skin is a major target of the TGF-β superfamily. Additionally, the relevance of skin carcinogenesis to carcinogenesis in other epithelial tissues and the accessibility of the skin make the epidermis an ideal tissue to study the role of TGF-β signaling in epithelial carcinogenesis. Glick et al.(13) reported that skin tumors devoid of TGF-β1 were associated with a high risk for malignant conversion. They also reported that grafts of v-rasHa-initiated TGF-β1-null keratinocytes progressed much more rapidly to SCCs than did wild-type TGF-β1 keratinocytes initiated with v-rasHa(14). Another advantage of using the skin is that the skin chemical carcinogenesis model is one of the most widely used and reliable models, which mimics the multistage nature of cancer development in humans. This system identified carcinogenesis as distinct stages of initiation, promotion, and malignant conversion (15). Initiation is induced by topical application of a subcarcinogenic dose of a particular carcinogen, e.g., DMBA. After initiation, a tumor promoter, e.g., TPA, is applied to elicit benign papillomas. After promotion, papillomas can persist, regress, or convert to malignancy (15). When this protocol was applied to transgenic mice expressing a dominant negative TGF-βRII in the epidermis, the transgenic mice exhibited acceleration of tumor formation and malignant progression (16, 17, 18). These experiments suggest that normal TGF-βRII functions are required for the tumor-suppressive effect of TGF-β. Cui et al.(19) performed chemical carcinogenesis experiments on transgenic mice, which overexpress TGF-β1 in the epidermis. TGF-β1 transgene inhibited benign tumor formation, yet enhanced progression to carcinomas. However, it is not known at which stage and by what mechanisms TGF-β1 switches its role from a tumor suppressor to a tumor promoter. Additionally, although TGF-β1 has been shown to promote tumor progression, no direct in vivo evidence has been reported documenting whether TGF-β1 directly induces distant metastasis. To address these questions, we have generated gene-switch TGF-β1 mice in which TGF-β1 transgene expression can be induced at specific stages of skin carcinogenesis (20). This system consists of two transgenic lines. The truncated ML promoter targets the transactivator GLVPc (a fusion molecule of the truncated progesterone receptor and the GAL4 DNA binding domain), and a tk promoter drives the TGF-β1 target gene (20). The ML vector targets transgene expression to all of the layers of the epidermis and through all of the stages of skin carcinogenesis including metastatic lesions (17). These two transgenic lines were mated to generate bigenic mice, and epidermal-specific expression of the TGF-β1 transgene is controlled by topical application of an antiprogestin (20). Here we provide in vivo evidence that TGF-β1 overexpression may directly induce metastasis by initiating events necessary for invasion, including loss of cell adhesion, early basement membrane degradation, and angiogenesis. Down-regulation of TGF-β signaling components in tumor epithelia may be a key event for the functional switch of TGF-β1 from a tumor suppressor to a metastasis promoter.

Skin Chemical Carcinogenesis.

Inducible and epidermal-specific TGF-β1 transgenic mice (ICR strain) were generated as described previously (20). A chemical carcinogenesis protocol was applied to 10-week-old mice as described previously (17). Briefly, 50 μg of DMBA was topically applied to the back skin of a mouse. After DMBA initiation (1 week), mice were topically treated with 5 μg of TPA, once/week for 20 weeks. TGF-β1 transgene expression was induced by topical application of a progesterone antagonist, ZK, to transgenic mouse skin tumors. ZK was given 10 μg/mouse (dissolved in 100 μl of EtOH), 3 times/week, beginning from 12 weeks after DMBA initiation.

Histological Analysis.

Tumor samples were fixed in 10% neutral-buffered formalin at 4°C overnight, embedded in paraffin, sectioned to 6-μm thickness, and stained with H&E. Tumor types were determined by histology following the criteria described previously (21).

Immunohistochemical Staining.

Immunohistochemistry was performed on formalin-fixed tumor sections as described previously (17). To stain for TGF-βRI and TGF-βRII, the sections were blocked with 5% normal goat serum at room temperature for 1 h followed by incubation with the antibody against TGF-βRI (V-22, 2 μg/ml diluted in 1% BSA-Tris-buffered saline; Santa Cruz Biotechnology, Santa Cruz, CA) or TGF-βRII (L-21, 4 μg/ml diluted in 1% BSA- Tris-buffered saline; Santa Cruz Biotechnology) at room temperature for 2 h. A secondary antibody, biotinylated donkey antigoat IgG (1:200; Santa Cruz Biotechnology) was then applied to the sections for 1 h at room temperature. Staining for Smad2, 3, and 4 was performed as described previously (22), using the Smad2/3 antibody (E-20, 1:100; Santa Cruz Biotechnology) that recognizes both Smad2 and Smad3, and the Smad4 antibody (C-20, 1:100; Santa Cruz Biotechnology). To stain cell adhesion molecules, the M.O.M. blocking reagents (Vector, Burlingame, CA) were applied to tumor sections before monoclonal antibodies for E-cadherin, β-catenin, and γ-catenin (all from Transduction Labs, San Diego, CA; 1:100) were applied to the sections at room temperature for 2 h. Biotinylated antimouse IgG (1:500; Vector) was then applied to the sections for 15 min at room temperature. The immune complex was detected by the avidin-biotin-peroxidase complex using Vectastain kits (Vector). Immune complexes were visualized with diaminobenzidine and counterstained with hematoxylin.

Immunofluorescence of BrdUrd Labeling and CD31 Staining.

Tumor-bearing mice were injected (i.p.) with BrdUrd 125 mg/kg in 0.9% NaCl. DMBA/TPA-induced tumors from bigenic TGF-β1 and control mice were dissected 1 h after the injection, fixed, and processed as described previously (18). Sections were incubated with FITC-conjugated monoclonal antibody to BrdUrd (PharMingen, San Diego, CA) and a guinea-pig antiserum to mouse K14, which reacts with the epithelial component of papillomas. K14 was visualized by biotinylated antiguinea pig IgG (Vector) and Streptavidin-Texas Red (Invitrogen, Carlsbad, CA). BrdUrd labeled cells were counted under 10× ocular and 20× objective lenses. Consecutive fields were counted throughout the entire tissue section. Tumor samples (6–10) were analyzed in each group. The labeling index was expressed as the mean of BrdUrd-positive cells/mm basement membrane ± SD. Immunofluorescence staining and quantitative analysis of vascular density were performed as described previously (17) using a rat antimouse CD31 (PECAM-1) antibody (PharMingen). Seven tumor samples were analyzed in each group. The percentage of stromal area covered by vessels was measured by computer analysis as described previously (17) and expressed as mean ± SD.

RPAs.

Total RNA was isolated from chemically induced tumors and the epidermis with RNAzol B (Tel-Test, Inc., Woodland, TX) as described previously (23). RPAs were performed using the RPA II kit (Ambion, Austin, TX) and 32P-labeled riboprobes. TSP-1-specific riboprobe was generated as described previously (17). The plasmid containing BamHI/SmaI fragment of mouse TSP-2 cDNA (corresponding to 1755–2065 bp; a gift from Dr. Jack Lawler) was used to label the TSP-2 riboprobe. The SmaI/NcoI fragment of a PCR product, which encompasses 634–886 bp of MMP2 cDNA, was subcloned into the pGem5 vector (Promega, Madison, WI). A PCR fragment, which encompasses 151–442 bp of MMP9 cDNA was subcloned into the pGem-T vector (Promega). The above plasmids were linearized at the 5′ end and served as templates for labeling 32P riboprobes. The intensity of a protected band in RPA was determined by densitometeric scanning of X-ray films.

In Situ Hybridization.

Tumors were frozen in OCT (Miles, Elkhart, IN) and sectioned to 10 μm on ProbeOn Plus slides (Fisher, Pittsburgh, PA). In situ hybridization was performed as described previously (22). Antisense and sense riboprobes specific for MMP2 and MMP9 were labeled with digoxigenin-11-UTP using the same templates used in PRA described above.

TGF-β1 Transgene Induction during the Middle Stage of Tumorigenesis Reveals a Role Switch from Tumor Suppression to Tumor Promotion.

To achieve epidermal-specific induction of TGF-β1 transgene expression, the ML vector was used to drive the transactivator GLVPc, and the constitutively active form of TGF-β1 was targeted by the tk promoter (20). A chemical carcinogenesis protocol was applied to bigenic mice (ML.GLVPc/tk.TGF-β1 double positive) and their nontransgenic or monogenic (ML.GLVPc or tk.TGF-β1) control mice (30 mice in each group). TGF-β1 transgene expression was induced by topical application of a progesterone antagonist, ZK, to bigenic mouse skin, beginning at 12 weeks after DMBA initiation, when chemically induced skin tumors began to arise (Fig. 1, A and B). We reasoned that the tumor-suppressive effect of TGF-β1 may be dominant before this stage, whereas its functional switch may occur after midstage TPA promotion. Bigenic mice treated with EtOH and monogenic mice treated with ZK exhibited tumor kinetics essentially the same as in nontransgenic mice with or without ZK treatment. Thus, these mice, without additional discrimination, served as controls for bigenic mice treated with ZK (TGF-β1 transgene induction). In control mice, tumor numbers continuously increased until the end of TPA promotion (Fig. 1,B). In contrast, ZK-treated bigenic mice exhibited a reduced tumor number (Fig. 1,B; P = 0.02), indicating that TGF-β1 induction at this stage still provides a tumor-suppressive effect. However, after termination of TPA promotion, control tumors began to regress, whereas ZK-treated bigenic tumors did not regress. Instead, they converted to malignant carcinomas at a much higher rate than control tumors (Fig. 1,C). TGF-β1 transgenic papillomas converted to carcinomas as early as 27 weeks after DMBA initiation. Bigenic mice (64%) treated with ZK developed carcinomas by 30 weeks after DMBA initiation and 81% of mice developed carcinomas by 40 weeks (Fig. 1,C), whereas only 5% of control mice developed carcinomas by 30 weeks, and 42% by 50 weeks (Fig. 1,C). Furthermore, TGF-β1 transgenic tumors metastasized to inguinal lymph nodes as early as 27 weeks after DMBA initiation, and 40% of TGF-β1 transgenic mice had already developed metastases by 40 weeks. Surprisingly, metastatic tumors (Fig. 2,B) developed from well-differentiated SCCs (Fig. 2,A). Higher magnification of the tumor sections revealed breakage of the basement membrane in focal areas (Fig. 2,C). Metastatic tumor cells exhibited expression of the epithelial marker, K14, and a marker for early malignant progression, K13 (data not shown). In contrast, <5% of control mice developed metastases from late-stage SCCs by 50 weeks after DMBA initiation (Fig. 2,D), which typically exhibited a poorly differentiated histotype (Fig. 2, D and E) and did not express K13 (not shown).

TGF-β1-resistant Papillomas Have Escaped from TGF-β1-induced Growth Arrest and Exhibit Down-Regulation of TGF-β Signaling Components.

The tumor kinetics of TGFβ1 transgenic tumors suggests that TGF-β1 transgenic papillomas responded to TGF-β1-induced growth inhibition at early stages but escaped such inhibition at later stages. To confirm this, in vivo BrdUrd labeling was performed on transgenic and control tumors. Similar to our previous observation that TGF-β1 transgene induction suppresses TPAinduced hyperproliferation after acute TPA application (20), TGF-β1 transgenic papillomas dissected 3 weeks after TGF-β1 transgene induction exhibited almost a 2-fold reduction in BrdUrd labeling index (Fig. 3; 39 ± 9.7 nuclei/mm basement membrane; n = 9) in comparison with control papillomas (Fig. 3; 69 ± 16 nuclei/mm basement membrane; n = 10; P < 0.001). In contrast, TGF-β1 transgenic papillomas dissected 15 weeks after TGF-β1 transgene induction (i.e., 27 weeks after DMBA initiation) exhibited a slightly higher BrdUrd labeling index (Fig. 3; 65 ± 11 nuclei/mm basement membrane; n = 8; P < 0.01) than control papillomas (Fig. 3; 56 ± 7.5 nuclei/mm basement membrane; n = 6; P < 0.05).

Studies have shown that loss of TGF-β signaling, as a result of mutations, decreased transcription or loss of proteins of TGF-βR or Smads, abrogates TGF-β1-induced growth inhibition (10, 11). Because mutations in TGF-βR or Smads have not been detected in chemically induced skin tumors (17, 22), we examined expression of TGF-β signaling components in TGF-β1 transgenic and control tumors at both the RNA and protein levels. Control and TGF-β1 transgenic papillomas 15 weeks after DMBA initiation (3 weeks after TGF-β1 transgene induction) expressed TGF-β signaling components, including TGF-βRI and TGF-βRII, Smad2, 3 and 4, at levels comparable with normal epidermis (data not shown). Control papillomas dissected 27 weeks after DMBA initiation exhibited reduced TGF-βRI staining in focal areas and uniform reduction in TGF-βRII staining but still expressed normal levels of Smad2, 3, and 4 in comparison with the epidermis (Fig. 4). Similar to control tumors, TGF-β1 transgenic papillomas dissected at the same stage (i.e., 15 weeks after TGF-β1 transgene induction) also showed a reduction in TGF-βRII staining (Fig. 4). Additionally, TGF-βRI, Smad2, 3, and 4 decreased to a barely detectable level in TGF-β1 transgenic papillomas (Fig. 4), although their expression at the mRNA level was comparable with control tumors (data not shown).

TGF-β1 Transgenic Papillomas Exhibit Earlier Loss of Membrane-associated E-Cadherin/Catenin Complex.

To determine the mechanisms of TGF-β1 transgene-induced early metastasis, we examined expression of E-cadherin/catenin, the complex which is required for epithelial cell adhesion and invasion suppression (24, 25). TGF-β1 transgenic papillomas dissected 3 weeks after TGF-β1 transgene induction exhibited strong E-cadherin staining in the cell membrane (data not shown). However, TGF-β1 transgenic papillomas dissected at 15 weeks after TGF-β1 transgene induction lost membrane-associated E-cadherin, which was redistributed in the cytoplasm (Fig. 5). Consistent with previous reports that E-cadherin is frequently lost in poorly differentiated malignancies (25), control mice retained E-cadherin in the membrane of benign papilloma cells at all of the stages but redistributed it to the cytoplasm in SCC cells (Fig. 5).

The cytoplasmic domain of E-cadherin interacts with β- and γ-catenins (25). To determine whether loss of membrane-associated E-cadherin is accompanied by the loss of membrane-associated catenins, we examined localization of β- and γ-catenins. In control tumors, both β- and γ-catenins stained strongly in the cell membrane of benign papillomas and differentiated cells of SCCs (Fig. 5). However, when TGF-β1 transgenic papillomas have lost E-cadherin in the membrane, they have also lost β- and γ-catenins in the membrane, which were redistributed to the cytoplasm (Fig. 5). Nuclear staining for β- and γ-catenins was not evident in TGF-β1 transgenic papillomas (Fig. 5).

TGF-β1 Transgenic Tumors Exhibit Increased Expression of MMP2 and MMP9.

Degradation of the basement membrane is critical for tumor invasion (26). MMP2 and MMP9 degrade type IV collagen, a major component of the basement membrane (26). Because TPA transiently induces MMP expression (27, 28), to determine the effect of TGF-β1 transgene expression we examined tumors 7 weeks after termination of TPA application (i.e., 15 weeks after TGF-β1 transgene induction). In situ hybridization showed that control papillomas expressed detectable levels of MMP2 (Fig. 6,A) and MMP9 (Fig. 6,B) in tumor cells and stromal cells adjacent to the basement membrane. However, TGF-β1 transgenic papillomas exhibited much stronger staining for MMP2 (Fig. 6,A) and MMP9 (Fig. 6 B) throughout all of the layers of tumor cells and stromal cells both adjacent to and away from the basement membrane.

To additionally quantitate changes in MMP expression in TGF-β1 transgenic tumors, RPAs were performed. Expression of MMP2 and MMP9 was detected at low levels in adult epidermis (Fig. 6,C). Control papillomas exhibited MMP2 or MMP9 expression ∼2-fold of those in the epidermis (Fig. 6, C and D; P < 0.05; n = 8). TGF-β1 transgenic papillomas exhibited an additional increase in MMP2 and MMP9 expression, ∼5-fold higher than normal epidermis (Fig. 6, C and D; P < 0.001; n = 7).

TGF-β1 Transgenic Tumors Exhibit an Increase in Angiogenesis.

Because TGF-β1 transgenic tumors exhibited early distant metastasis, we examined angiogenesis in these tumors dissected 15 weeks after TGF-β1 transgene induction. CD31, a marker of endothelial intercellular junctions as well as platelets and leukocyte subsets (29), was stained by immunofluorescence. TGF-β1 transgenic papillomas exhibited a 5-fold increase in angiogenesis, in comparison with control papillomas (Fig. 7,A). The percentage of the stromal area covered by vessels in TGF-β1 papillomas increased to 64% ± 23% (n = 7) compared with 13% ± 10% in control papillomas (Fig. 7 B; n = 7; P < 0.01).

Consistent with the increased angiogenesis in TGF-β1 transgenic tumors, two angiogenesis inhibitors, TSP-1 and TSP-2, exhibited ∼50% reduction in TGF-β1 transgenic papillomas in comparison with their corresponding control tumors (Fig. 7, C and D).

Down-Regulation of TGF-β Signaling in Tumor Epithelia May Be a Key Event to Switch the Role of TGF-β1 from a Tumor Suppressor to a Tumor Promoter.

We observed a functional switch of TGF-β1 from a tumor suppressor to a tumor promoter, which correlated with decreased levels of TGF-β signaling components in TGF-β1 transgenic papillomas. Previously, we have shown that Smads were lost in chemically induced skin SCCs at a late stage (22). Therefore, earlier down-regulation of TGF-β signaling components in TGF-β1 transgenic tumors may play a key role in abrogation of TGF-β1-induced growth arrest in tumor cells. It has been observed from various cancer cells that deregulation of one of the TGF-β signaling components, either the receptors or Smads, is sufficient to abolish TGF-β-induced growth inhibition, and transfection with the corresponding wild-type component in these tumor cells restored TGF-β-induced growth arrest (10). However, in TGF-β1 transgenic tumors, deregulation of a single TGF-β signaling component may not be sufficient to abolish TGF-β-induced growth arrest, because TGF-β1 transgene expression forces TGF-β signaling. Therefore, only the tumors that have lost multiple TGF-β signaling components may be able to escape from TGF-β1 transgene-induced growth inhibition and progress to a later stage. Because both TGF-β1 transgenic epidermis and the earlier stage TGF-β1 transgenic papillomas exhibited normal levels of TGF-β signaling components, down-regulation of TGF-β signaling components is not a direct effect of TGF-β1 transgene expression. It is likely that TGF-β1 transgene expression has inhibited growth of tumors with normal levels of TGF-β signaling components and selected TGF-β-resistant papillomas that have reduced TGF-β signaling components. It is also possible that TGF-β1 transgene expression, in cooperation with other genetic alterations or increased protease activity in late stage papillomas, contributes to the increased degradation of TGF-β signaling components.

Once tumor cells have escaped from TGF-β1-induced growth arrest, the tumor promotion effect of TGF-β1 apparently becomes dominant. The question remains how TGF-β1 promotes tumor invasion if the signaling components are significantly reduced. One possible explanation is that TGF-β1-induced growth arrest may require a higher level of TGF-β signaling components than other TGF-β1 functions. Thus, although proteins of TGF-β signaling components were significantly decreased in TGF-β1 transgenic papillomas, they may be sufficient to mediate TGF-β1-induced tumor invasion. Supporting this hypothesis, expression of a low level of the dominant-negative TGF-βRII only blocked TGF-β-induced growth arrest but not TGF-β-induced transcriptional regulation of extracellular matrix proteins (30). Secondly, reduced Smad signaling may primarily affect TGF-β-induced growth inhibition, whereas alternative TGF-β signaling pathways may be required for TGF-β-induced tumor invasion. Supporting this hypothesis, in vitro studies have shown that TGF-β1-induced epithelial-mesenchymal transition in mammary epithelial cells requires functions of phosphatidylinositol 3-kinase (31) as well as RhoA activation (32); and blocking Smad signaling only abolishes TGF-β-induced growth inhibition but not epithelial-mesenchymal transition (32). Lastly, TGF-β1 may also induce tumor invasion via its paracrine effect on tumor stroma, which does not depend on levels of TGF-β signaling components in tumor epithelia. This is evident because TGF-β1 transgene induction increased MMP expression not only in tumor epithelia but also in stromal cells (Fig. 6).

TGF-β1 Transgene Expression Initiates Events Required for Tumor Invasion and Distant Metastasis.

The first requirement for tumor invasion to take place is a decrease in tumor cell adhesion. The E-cadherin/catenin complex plays a dominant role in epithelial cell adhesion and is also considered as an invasion suppressor (24, 25). Increasing evidence suggests that disruption of the E-cadherin/catenin complex is the cause, rather than the consequence of tumor invasion (24, 25). In the current study, we found that TGF-β1 reduces membrane-associated E-cadherin much earlier than the actual malignant progression (i.e., in benign papillomas). Similarly, β- and γ-catenins were also redistributed from the cell membrane to the cytoplasm in TGF-β1 transgenic papillomas. Because nuclear staining of β- and γ-catenins was not evident in TGF-β1 transgenic papillomas, redistribution of E-cadherin/catenin complex induced by TGF-β1 must primarily affect cell adhesion to facilitate tumor invasion.

Once cell motility is changed by loss of adhesion molecules, the next step necessary for tumor invasion is the degradation of the basement membrane. Although either the tumor epithelia or stromal cells in many cancer types can produce MMPs for basement membrane degradation, it is not clear which cell types express MMP2 and MMP9 during skin chemical carcinogenesis. In the present study, we observed that expression of MMP2 and MMP9 in nontransgenic papillomas was located in both the tumor cells and the adjacent stromal fibroblasts (Fig. 6). This result is in contrast to a recent report demonstrating that MMP9 is elevated exclusively in bone marrow-derived cells in the tumor stroma in human papillomavirus 16-induced skin carcinogenesis (33). Because we dissected tumors 7 weeks after termination of TPA application, increased MMP expression is unlikely a direct effect of TPA. Nevertheless, chemical carcinogenesis may activate MMP2 and MMP9 expression using mechanisms different from that in human papillomavirus 16 transgenic tumors. For instance, ∼90% of chemically induced skin tumors harbor c-rasHa mutations, and they exhibit elevation in the expression of epidermal growth factor family members (34). Both events have been shown to stimulate MMP expression in keratinocytes (35, 36, 37). TGF-β1 transgenic papillomas exhibited an additional increase in MMP2 and MMP9 expression in both tumor cells and stromal fibroblasts (Fig. 5), suggesting that TGF-β1 exerts both autocrine and paracrine effects on MMP expression. Because we only observed elevated expression of MMP2 and MMP9 in tumors (Fig. 6) but not in TGF-β1 transgenic epidermis (data not shown), elevation of MMP2 and MMP9 may require other oncogenic events in addition to TGF-β1 induction.

Loss of cell adhesion and basement membrane degradation are required to initiate local invasion, whereas increased angiogenesis is essential for autonomous tumor growth and distant metastasis. We have shown that TGF-β1 transgene expression induces angiogenesis in normal skin (20). However, during skin carcinogenesis, TGF-β1 transgene expression additionally increased angiogenesis, possibly in cooperation with other oncogenic events. For instance, TGF-β1 transgenic skin did not exhibit reduced expression of TSP-1 and TSP-2 (data not shown), two potent angiogenesis inhibitors (38, 39), whereas TGF-β1 transgenic papillomas exhibited decreased mRNA levels of TSP-1 and TSP-2 compared with control tumors. It has been recently reported that Smad4 directly induces TSP-1 expression (40). Therefore, it is possible that TGF-β1 transgene may not be able to reduce TSP-1 expression when the epidermis or early stage papillomas still retain a high level of Smad4. When Smads are down-regulated in TGF-β1 transgenic papillomas, the TGF-β1 transgene may act either by itself or in cooperation with other oncogenic events to reduce TSP expression. Both TSP-1 and TSP-2 have been shown recently to directly inhibit activation of MMP2 and MMP9 via protein-protein interactions (41, 42). Therefore, decreased TSP-1 and TSP-2 may also contribute to increased basement membrane degradation in TGF-β1 transgenic tumors. Because MMP2 and MMP9 also increase angiogenesis (43, 44), inhibition of their activation may be an important antiangiogenic mechanism of TSP-1 or TSP-2.

In summary, we have provided in vivo evidence that TGF-β1 overexpression, in concert with decreased TGF-β signaling components in tumor epithelia, promotes metastasis. Our study demonstrates that TGF-β1 exerts autocrine effects on tumor epithelia (e.g., redistribution of cell adhesion molecules and increased MMP expression in tumor cells) and paracrine effects on tumor stroma (e.g., increased MMP expression in tumor stroma and increased angiogenesis). TGF-β1-induced molecular changes involved in malignant progression could occur either at the transcriptional level (e.g., changes in expression of MMPs and TSPs) or at the protein level (e.g., changes in TGF-β signaling components and adhesion molecules). Detailed molecular mechanisms by which TGF-β1 regulates these molecules remain to be explored. We should also point out that many of these events (e.g., loss of cell adhesion, increase in MMPs, and reduced TSPs) are not elicited in TGF-β1 transgenic skin or early stage papillomas. Therefore, either tumor cells responded differently to TGF-β1 than normal cells, or cooperation of TGF-β1 with other oncogenic events is required for eliciting these invasive events. Our study also suggests that histologically appearing well-differentiated epithelial tumors may have a much poorer prognosis if TGF-β1 is overexpressed in concert with decreased TGF-β signaling components in tumor epithelia. Given that TGF-β1 overexpression is detected frequently in malignant human tumors, screening TGF-β1 expression levels and its downstream signaling components may have an important impact on predicting prognosis and therapeutic approaches.

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.

1

This work was supported by NIH Grant CA79998 (to X-J. W).

4

The abbreviations used are: TGF, transforming growth factor; SCC, squamous cell carcinoma; DMBA, dimethylbenzanthracene; TPA, 12-O-tetradecanoylphorbol-13acetate; ML, mouse loricrin; tk, thymidine kinase; ZK, ZK98.734; BrdUrd, bromodeoxyuridine; K, keratin; RPA, RNase protection assay; TSP, thrombospondin; MMP, matrix metalloproteinase; EtOH, ethanol.

We thank Dr. Jack Lawler (Beth Israel Deaconess Medical Center, Boston, MA) for providing mouse TSP-2 cDNA template. We also thank Drs. Dennis R. Roop and Xin-Hua Feng for their critical comments on the manuscript. Mattie Brooks and Donna Wang provided excellent technical assistance.

1
Bottinger E. P., Letterio J. J., Roberts A. B. Biology of TGF-β in knockout and transgenic mouse models.
Kidney Int.
,
51
:
1355
-1360,  
1997
.
2
Pepper M. S. Transforming growth factor-β: vasculogenesis, angiogenesis, and vessel wall integrity.
Cytokine Growth Factor Rev.
,
8
:
21
-43,  
1997
.
3
Roberts A. B., McCune B. K., Sporn M. B. TGF-β: regulation of extracellular matrix.
Kidney Int.
,
41
:
557
-559,  
1992
.
4
Roberts A. B., Sporn M. B. Physiological actions and clinical applications of transforming growth factor-β (TGF-β).
Growth Factors
,
8
:
1
-9,  
1993
.
5
Sporn M. B., Roberts A. B. Transforming growth factor-β: recent progress and new challenges.
J. Cell Biol.
,
119
:
1017
-1021,  
1992
.
6
Letterio J. J., Roberts A. B. TGF-β: a critical modulator of immune cell function.
Clin. Immunol. Immunopathol.
,
84
:
244
-250,  
1997
.
7
Massague J., Blain S. W., Lo R. S. TGFβ signaling in growth control, cancer, and heritable disorders.
Cell
,
103
:
295
-309,  
2000
.
8
Kavsak P., Rasmussen R. K., Causing C. G., Bonni S., Zhu H., Thomsen G. H., Wrana J. L. Smad7 binds to smurf2 to form an E3 ubiquitin ligase that targets the TGFβ receptor for degradation.
Mol. Cell
,
6
:
1365
-1375,  
2000
.
9
Tang B., Bottinger E. P., Jakowlew S. B., Bagnall K. M., Mariano J., Anver M. R., Letterio J. J., Wakefield L. M. Transforming growth factor-β-1 is a new form of tumor suppressor with true haploid insufficiency.
Nat. Med.
,
4
:
802
-807,  
1998
.
10
Gold L. I. The role for transforming growth factor-β (TGF-β) in human cancer.
Crit. Rev. Oncog.
,
10
:
303
-360,  
1999
.
11
Reiss M. TGF-β and cancer.
Microbes. Infect.
,
1
:
1327
-1347,  
1999
.
12
Factor V. M., Kao C. Y., Santoni-Rugiu E., Woitach J. T., Jensen M. R., Thorgeirsson S. S. Constitutive expression of mature transforming growth factor β1 in the liver accelerates hepatocarcinogenesis in transgenic mice.
Cancer Res.
,
57
:
2089
-2095,  
1997
.
13
Glick A. B., Kulkarni A. B., Tennenbaum T., Hennings H., Flanders K. C., O’Reilly M., Sporn M. B., Karlsson S., Yuspa S. H. Loss of expression of transforming growth factor β in skin and skin tumors is associated with hyperproliferation and a high risk for malignant conversion.
Proc. Natl. Acad. Sci. USA
,
90
:
6076
-6080,  
1993
.
14
Glick A. B., Lee M. M., Darwiche N., Kulkarni A. B., Karlsson S., Yuspa S. H. Targeted deletion of the TGF-β 1 gene causes rapid progression to squamous cell carcinoma.
Genes Dev.
,
8
:
2429
-2440,  
1994
.
15
DiGiovanni J. Multistage carcinogenesis in mouse skin.
Pharmacol. Ther.
,
54
:
63
-128,  
1992
.
16
Amendt C., Schirmacher P., Weber H., Blessing M. Expression of a dominant negative type II TGF-β receptor in mouse skin results in an increase in carcinoma incidence and an acceleration of carcinoma development.
Oncogene
,
17
:
25
-34,  
1998
.
17
Go C., Li P., Wang X. J. Blocking transforming growth factor β signaling in transgenic epidermis accelerates chemical carcinogenesis: a mechanism associated with increased angiogenesis.
Cancer Res.
,
59
:
2861
-2868,  
1999
.
18
Go C., He W., Zhong L., Li P., Huang J., Brinkley B. R., Wang X. J. Aberrant cell cycle progression contributes to the early-stage accelerated carcinogenesis in transgenic epidermis expressing the dominant negative TGFβRII.
Oncogene
,
19
:
3623
-3631,  
2000
.
19
Cui W., Fowlis D. J., Bryson S., Duffie E., Ireland H., Balmain A., Akhurst R. J. TGFβ1 inhibits the formation of benign skin tumors, but enhances progression to invasive spindle carcinomas in transgenic mice.
Cell
,
86
:
531
-542,  
1996
.
20
Wang X. J., Liefer K. M., Tsai S. Y., O’Malley B. W., Roop D. R. Development of gene-switch transgenic mice that inducibly express transforming growth factor β1 in the epidermis.
Proc. Natl. Acad. Sci. USA
,
96
:
8483
-8488,  
1999
.
21
Aldaz C. M., Conti C. J., Klein-Szanto A. J., Slaga T. J. Progressive dysplasia and aneuploidy are hallmarks of mouse skin papillomas: relevance to malignancy.
Proc. Natl. Acad. Sci. USA
,
84
:
2029
-2032,  
1987
.
22
He W., Cao T., Smith D. A., Myers T. E., Wang X. J. Smads mediate signaling of the TGFβ superfamily in normal keratinocytes but are lost during skin chemical carcinogenesis.
Oncogene
,
20
:
471
-483,  
2001
.
23
Wang X. J., Greenhalgh D. A., Jiang A. B., He D. C., Zhong L., Medina D., Brinkley B. R., Roop D. R. Expression of a p53 mutant in the epidermis of transgenic mice accelerates chemical carcinogenesis.
Oncogene
,
17
:
35
-45,  
1998
.
24
Vermeulen S., Van M. V, Van Hoorde L., Van Roy F., Bracke M., Mareel M. Regulation of the invasion suppressor function of the cadherin/catenin complex.
Pathol. Res. Pract.
,
192
:
694
-707,  
1996
.
25
Smith M. E., Pignatelli M. The molecular histology of neoplasia: the role of the cadherin/catenin complex.
Histopathology
,
31
:
107
-111,  
1997
.
26
MacDougall J. R., Matrisian L. M. Contributions of tumor and stromal matrix metalloproteinases to tumor progression, invasion and metastasis.
Cancer Metastasis Rev.
,
14
:
351
-362,  
1995
.
27
Sato H., Seiki M. Regulatory mechanism of 92 kDa type IV collagenase gene expression which is associated with invasiveness of tumor cells.
Oncogene
,
8
:
395
-405,  
1993
.
28
Mackay A. R., Ballin M., Pelina M. D., Farina A. R., Nason A. M., Hartzler J. L., Thorgeirsson U. P. Effect of phorbol ester and cytokines on matrix metalloproteinase and tissue inhibitor of metalloproteinase expression in tumor and normal cell lines.
Invasion Metastasis
,
12
:
168
-184,  
1992
.
29
Charpin C., Devictor B., Bergeret D., Andrac L., Boulat J., Horschowski N., Lavaut M. N., Piana L. CD31 quantitative immunocytochemical assays in breast carcinomas. Correlation with current prognostic factors.
Am. J. Clin. Pathol.
,
103
:
443
-448,  
1995
.
30
Chen R. H., Ebner R., Derynck R. Inactivation of the type II receptor reveals two receptor pathways for the diverse TGF-β activities.
Science (Wash. DC)
,
260
:
1335
-1338,  
1993
.
31
Bakin A. V., Tomlinson A. K., Bhowmick N. A., Moses H. L., Arteaga C. L. Phosphatidylinositol 3-kinase function is required for transforming growth factor β-mediated epithelial to mesenchymal transition and cell migration.
J. Biol. Chem.
,
275
:
36803
-36810,  
2000
.
32
Bhowmick N. A., Ghiassi M., Bakin A., Aakre M., Lundquist C. A., Engel M. E., Arteaga C. L., Moses H. L. Transforming growth factor-β1 mediates epithelial to mesenchymal transdifferentiation through a RhoA-dependent mechanism.
Mol. Biol. Cell
,
12
:
27
-36,  
2001
.
33
Coussens L. M., Tinkle C. L., Hanahan D., Werb Z. MMP-9 supplied by bone marrow-derived cells contributes to skin carcinogenesis.
Cell
,
103
:
481
-490,  
2000
.
34
Rho O., Beltran L. M., Gimenez-Conti I. B., DiGiovanni J. Altered expression of the epidermal growth factor receptor and transforming growth factor-α during multistage skin carcinogenesis in SENCAR mice.
Mol. Carcinog.
,
11
:
19
-28,  
1994
.
35
Meade-Tollin L. C., Boukamp P., Fusenig N. E., Bowen C. P., Tsang T. C., Bowden G. T. Differential expression of matrix metalloproteinases in activated c-ras-Ha-transfected immortalized human keratinocytes.
Br. J. Cancer
,
77
:
724
-730,  
1998
.
36
Sato H., Kita M., Seiki M. v-Src activates the expression of 92-kDa type IV collagenase gene through the AP-1 site and the GT box homologous to retinoblastoma control elements. A mechanism regulating gene expression independent of that by inflammatory cytokines.
J. Biol. Chem.
,
268
:
23460
-23468,  
1993
.
37
Charvat S., Le Griel C., Chignol M. C., Schmitt D., Serres M. Ras-transfection up-regulated HaCaT cell migration: inhibition by Marimastat.
Clin. Exp. Metastasis
,
17
:
677
-685,  
1999
.
38
Streit M., Riccardi L., Velasco P., Brown L. F., Hawighorst T., Bornstein P., Detmar M. Thrombospondin-2: a potent endogenous inhibitor of tumor growth and angiogenesis.
Proc. Natl. Acad. Sci. USA
,
96
:
14888
-14893,  
1999
.
39
Streit M., Velasco P., Brown L. F., Skobe M., Richard L., Riccardi L., Lawler J., Detmar M. Overexpression of thrombospondin-1 decreases angiogenesis and inhibits the growth of human cutaneous squamous cell carcinomas.
Am. J. Pathol.
,
155
:
441
-452,  
1999
.
40
Schwarte-Waldhoff I., Volpert O. V., Bouck N. P., Sipos B., Hahn S. A., Klein-Scory S., Luttges J., Kloppel G., Graeven U., Eilert-Micus C., Hintelmann A., Schmiegel W. Smad4/DPC4-mediated tumor suppression through suppression of angiogenesis.
Proc. Natl. Acad. Sci. USA
,
97
:
9624
-9629,  
2000
.
41
Bein K., Simons M. Thrombospondin type 1 repeats interact with matrix metalloproteinase 2. Regulation of metalloproteinase activity.
J. Biol. Chem.
,
275
:
32167
-32173,  
2000
.
42
Yang Z., Kyriakides T. R., Bornstein P. Matricellular proteins as modulators of cell-matrix interactions: adhesive defect in thrombospondin 2-null fibroblasts is a consequence of increased levels of matrix metalloproteinase-2.
Mol. Biol. Cell
,
11
:
3353
-3364,  
2000
.
43
Bergers G., Brekken R., McMahon G., Vu T. H., Itoh T., Tamaki K., Tanzawa K., Thorpe P., Itohara S., Werb Z., Hanahan D. Matrix metalloproteinase-9 triggers the angiogenic switch during carcinogenesis.
Nat. Cell Biol.
,
2
:
737
-744,  
2000
.
44
Itoh T., Tanioka M., Yoshida H., Yoshioka T., Nishimoto H., Itohara S. Reduced angiogenesis and tumor progression in gelatinase A-deficient mice.
Cancer Res.
,
58
:
1048
-1051,  
1998
.