Conditional Overexpression of Active Transforming Growth Factor β1 In vivo Accelerates Metastases of Transgenic Mammary Tumors

Multiple cellular mechanisms control cancer progression and metastases, including tumor cell proliferation, survival, motility, and invasion, as well as extracellular matrix components, neovascularization, and host immunosuppression. The cytokine transforming growth factor β (TGF-β) can contribute to each of these processes, placing TGF-β at a pivotal position to regulate tumor progression. TGF-βs are members of a large superfamily of pleiotropic cytokines that includes the activins and bone morphogenetic proteins (BMPs; ref. 1). Members of the TGF-β family regulate complex processes such as cell proliferation, differentiation, adhesion, cell-cell and cell-matrix interactions, motility, and cell death. TGF-βs bind to a heteromeric complex of serine/threonine kinases, the type I and type II receptors (TβRI and TβRII) (2). Following ligand binding to TβRII, TβRI is recruited to the complex, which allows for the constitutively active TβRII kinase to transphosphorylate and activate TβRI (3). TβRI, in turn, phosphorylates Smad2 and Smad3, which then associate with Smad4 and translocate to the nucleus where they regulate gene transcription (2, 4). In addition to Smads, other signaling pathways more recently have been implicated in TGF-β actions. These include the extracellular signal-regulated kinase (ERK), c-Jun NH₂-terminal kinase (JNK), p38 mitogen-activated protein kinase (MAPK), phosphatidylinositol-3′ kinase (PI3K), and Rho GTPases (reviewed in refs. 4, 5). Overall, the critical role of these non-Smad pathways on mediating the cellular effects of TGF-β remains to be fully characterized.

It generally is accepted that TGF-β can behave as a tumor suppressor and tumor promoter. Its tumor suppressor role can be explained by its ability to inhibit cell proliferation, maintain tissue architecture (6), inhibit genomic instability (7), and induce replicative senescence and apoptosis (8). Overexpression of active TGF-β under the control of tissue-specific promoters in transgenic mice can delay or protect from carcinogen- or oncogene-induced carcinomas (9, 10). Furthermore, mice with complete or partial disruption of Tgfb1 or Smad genes are prone to the development of carcinomas (6, 11, 12). Attenuation of autocrine TGF-β signaling by expression of a dominant-negative TβRII results in accelerated lobuloalveolar mammary development (13), enhanced propensity for carcinogen-induced lung, mammary, and skin tumors (14, 15), and spontaneous invasive mammary carcinomas (16). Finally, mutations in the TGFBR2 gene occur in sporadic and inherited colon cancers with microsatellite instability (17), and restoration of TβRII by transfection reverses transformation in certain colon cancer cell lines (18). Although these studies support the tumor-suppressive role of endogenous TGF-β, it should be noted that to date administration of exogenous TGF-β has not been shown to inhibit established cancers.

Conversely, there is increasing evidence to indicate that high production and/or activation of TGF-β in tumors can foster cancer progression by autocrine and/or paracrine mechanisms (reviewed in refs. 5,19, 20). Overexpression of TGF-β ligands has been reported in most cancers (reviewed in ref. 21). These high TGF-β levels in tumor tissues correlate with markers of a more metastatic phenotype and/or poor patient outcome, and many tumor cells exhibit increased invasiveness in response to TGF-β (reviewed in ref. 22). TGF-β also can induce an epithelial-to-mesenchymal transition in tumor and nontumor epithelial cells (23, 24). Reexpression of TβRII in colon cancer cells with low invasive potential restores tumor cell invasiveness (25), and induction of TGF-β1 in papillomas rapidly induces metastatic carcinomas (26). Forced expression of dominant active Smad2 in squamous cancer cells also results in enhanced tumor cell motility and metastatic dissemination (27). Further underscoring the tumor-promoting role of autocrine TGF-β, expression of dominant-negative TβRII in metastatic cancer cells prevents epithelial-to-mesenchymal transition and inhibits motility, tumorigenicity, and metastases (28). These data suggest that TGF-β may select for more metastatic cancers. Mice overexpressing active TGF-β1 in suprabasal keratinocytes develop fewer benign papillomas compared with controls. However, once tumors develop, the transgenic tumors rapidly acquire a spindle cell phenotype, overexpress TGF-β3, and metastasize (10). More recently, overexpression of active TGF-β1 or activated TβRI in the mammary gland of transgenic mice accelerated metastases derived from neu-induced primary mammary tumors (29, 30) Finally, colon cancers with inactivating mutations of the TGFBR2 gene exhibit favorable survival compared with TβRII-positive colon cancers (31). These observations suggest that loss of autocrine TGF-β signaling in carcinomas may limit systemic metastases.

The variable effects on tumorigenesis of an excess of TGF-β in transgenic cancer models suggest that the net effect of TGF-β on tumor progression may well depend on the timing and context during stochastic transformation in which this overexpression occurs. Therefore, we have developed a triple transgenic model of oncogene-induced mammary carcinoma in which active TGF-β1 can be conditionally regulated, thus avoiding the potential tumor-suppressive effects during early phases of transformation. In this model, a short induction of TGF-β after primary tumors were established clearly accelerated metastatic progression at least in part by a direct effect on tumor cells.

Mammary tumors from one mouse mammary tumor virus (MMTV)-polyomavirus middle T antigen (PyVmT) female mouse (32) were digested at 37°C for 4 hours in 3 mg/mL collagenase A (Sigma, St. Louis, MO) in PBS (pH 7.4). The cell suspension was plated on Growth Factor-Reduced Matrigel (BD Biosciences Pharmingen, San Diego, CA) in Dulbecco’s Modified Eagle’s Medium/F12 (50:50; Life Technologies, Rockville, MD), 10% fetal calf serum (FCS) and 50 ng/mL insulin (Cambrex, East Rutherford, NJ) and cultured at 37°C in 5% CO₂. After eight passages, cells underwent crisis. Cells were clonally isolated, and the resulting colonies were pooled and termed PyVmT cells. All of the experiments used cells between passage 18 and 25. When indicated, cells were serum deprived overnight (0.5% FCS) and cultured for 30 minutes in the presence of recombinant human TGF-β1 (2 ng/mL; R&D Systems, Minneapolis, MN), Fc:TβRII (20 nmol/L; provided by Phillip Gotwals, Biogen, Inc., Cambridge, MA), 20 μmol/L U0126 (Promega, Madison, WI), 20 μmol/L LY294002, or 20 μmol/L SB202190 (both from Calbiochem, San Diego, CA). To generate the PyVmT:Neo and PyVmT:αsT cells, retroviral particles encoded by the following vectors were used to infect PyVmT cells: pLα s-TGF-β1SN encoding antisense mouse TGF-β1 cDNA, or the empty parental vector pLXSN (33). Dr. Emmanuel Akporiaye (University of Arizona, Tucson, AZ) provided the vectors. Infected PyVmT cells were selected in 0.4 mg/mL Geneticin (G418; Omega Scientific, Tarzana, CA). All of the experiments were performed on pooled clones.

Cells were labeled with Sp-DiOC18(3) (Molecular Probes, Eugene, OR) and seeded in the upper chamber of transwells fitted with Matrigel-coated, 8-μm pore polycarbonate filters (Corning Inc. Life Sciences, Acton, MA). Lower chambers contained 2.5% serum with or without 20 nmol/L Fc:Tβ RII, 20 μmol/L U0126, 20 μmol/L LY294002, or 20 μmol/L SB202190. After 24 hours, cells were scraped from upper filter surfaces, and cells on the lower surfaces were photographed using fluorescence microscopy. Fluorescence was quantified using Scion Image (Frederick, MD) software.

Mammary glands, tumors, or cell lysates were harvested and homogenized as described previously (34). Fractionation of tumors into nuclear and cytosolic fractions was done as described by Lenferink et al. (35). Western analyses were performed as described previously (34) using the following antibodies: cytokeratin 8 (C51), α-actinin (H-300), Smad4 (B-8), Smad7 (H79), p38, c-jun, and tubulin (Santa Cruz Biotechnology, Santa Cruz, CA); total and P-MAPK (Promega); total and S473 P-Akt and P-Smad2 (Upstate Biotechnology, Lake Placid, NY); Smad 2/3 (BD Biosciences Pharmingen); P-p38 (Cell Signaling, Beverly, MA); and PyVmT (pAb 701; a gift from Dr. Steven Dilworth, Imperial Cancer Research Fund, London, UK). Total RNA was harvested from mammary glands using TRIzol (Invitrogen, Carlsbad, CA) according to manufacturer’s directions. Northern analysis using total RNA (20 μg) was performed as described (36).

Cells (2 × 10⁶) were cultured for 24 hours in 3 mL of serum-free media. Conditioned medium was collected, and TGF-β1, TGF-β2, and IFN-γ levels in it were determined using ELISA (each from R&D Systems). Mouse serum was tested directly in the TGF-β1 ELISA in the absence of acid activation.

PyVmT, PyVmT:Neo, and PyVmT:αsT cells (1.0 × 10⁶) were injected into no. 4 mammary glands of FVB virgin female mice via surgical exposure of the gland. Some mice were treated with 5 mg/kg/d Fc:TβRII by intraperitoneal injection starting on day after PyVmT tumor cell inoculation. Mice were monitored for tumor formation twice weekly by palpation, and tumor volume was calculated using the formula: volume = width² × length/2. At 100 days after initial tumor palpation, mice were sacrificed, and tissues were harvested.

Tissues were fixed in 10% formalin (VWR Scientific, West Chester, PA). Hematoxylin-stained whole mounts of right inguinal mammary glands were prepared as described previously (34). Sections of paraffin-embedded mammary glands (5 μm) were stained with H&E (all from Sigma). Detection of apoptosis by terminal deoxynucleotidyl transferase-mediated nick end labeling (TUNEL) analysis was performed using the Apoptag Detection Kit (Serologicals Corp., Norcross, GA). Immunohistochemistry was performed using an antibody against proliferating cell nuclear antigen (PCNA; Neomarkers, Freemont, CA) as described (36). Immunofluorescence for Smad2 and active TGF-β1 was performed using cryosections as described previously (37) using the following antibodies: active TGF-β1 (catalogue no. AF-101-NA, lot FS08; R&D Systems), Smad2/3 (Santa Cruz Biotechnology), and secondary antibodies labeled with Alexa-488 or Alexa-594 fluorochromes (Molecular Probes). Images were captured using a 12-bit charge-coupled device (KAF-1400; 1317 × 1035 6.8-mm² pixels) digital camera (Xillix, Vancouver, Canada).

The 1.02-kb constitutively active simian TGF-β1 cDNA fragment (38) was subcloned into the vector pTet-Splice (Life Technologies). The linearized transgene was injected into one-cell FVB mouse embryos. All of the resulting pups were screened for presence of the transgene using primer pairs within pTet-Splice or by Southern analysis using the rabbit β-globin intron sequence from the transgene (38). Two transgenic founders capable of transmitting the transgene to F1 mice were identified. Transgenic F1s (pure FVB) were bred with MMTV-reverse-tetracycline-transactivator (rtTA) mice (39) to generate MMTV-rtTA/TetOp-TGF-β1^S223/225 (r/T) mice. When indicated, drinking water was supplemented with 2 mg/mL doxycycline (DOX) in 5% sucrose. DOX was added to the drinking water of pregnant female mice at 7.5 days postcoitus and maintained through parturition and lactation. For tumor studies, double transgenic r/T female mice were crossed with MMTV-PyVmT mice. Only age-matched virgin female mice were analyzed.

A cell line was derived from an MMTV-PyVmT mouse mammary cancer. These cells were epithelial because they expressed keratin-8 and maintained expression of middle T (Fig. 1,A). Proliferation of PyVmT cells was unaltered by exogenous TGF-β1 (data not shown). In serum-starved PyVmT cells, low levels of the active (phosphorylated) Akt, p38, and MAPK were observed. Treatment of PyVmT cells with TGF-β1 resulted in a rapid increase in the levels of P-Akt and P-p38 but not P-MAPK (Fig. 1,A–C). Pretreatment of cells with TGF-β inhibitory fusion protein Fc:TβRII blocked ligand-induced phosphorylation of Akt and p38 (Fig. 1,A and B) and reduced basal P-Akt (Fig. 1,A) and P-MAPK (Fig. 1 C), suggesting that autocrine TGF-β signaling regulates the activation of Akt and MAPK in these cells.

We investigated the role of autocrine TGF-β signaling on motility and survival in PyVmT cells. Basal migration through transwells under low-serum conditions was variably reduced by the addition of Fc:TβRII>LY294002>SB202190>U0126 but enhanced by TGF-β1. Inhibition of PI3K and p38 with LY294002 and SB202190, respectively, completely blocked TGF-β-induced motility, whereas blockade of MAPK with U0126 did so partially (Fig. 1,D). Similar results were obtained in 16-hour wound closure assays in low serum and in the presence of a DNA synthesis-blocking dose of mitomycin C (not shown), suggesting that the observed effects on tumor cell migration were unlikely to be significantly related to changes in proliferation and/or survival. TUNEL analysis was used to measure cell survival. After 48 hours of serum starvation, treatment with Fc:TβRII increased the proportion of TUNEL-positive cells from 11.6 to 35.7%. Addition of TGF-β1 decreased the TUNEL-positive cells to 4%. Blockade of PI3K with LY294002 increased the rate of apoptosis in the absence and presence of TGF-β1 (51.6% and 49.6%, respectively), suggesting that PI3K signaling is required for TGF-β-induced cell survival. Blockade of the MAPK or p38 did not significantly alter basal or TGF-β-stimulated cell survival (Fig. 1 E).

To determine whether autocrine TGF-β was causally associated with tumor cell motility and survival, we stably transduced the PyVmT cells with a retrovirus encoding an antisense mouse TGF-β1 (PyVmT:αsT). PyVmT (parental) and pooled clones of PyVmT:Neo (vector control) cells produced >700 pg/mL/48 h TGF-β1 as measured by ELISA, whereas the αsT cells secreted 52 pg/mL/48 h (Fig. 2,A, left). Secretion of TGF-β2 or IFN-γ was similar in PyVmT, PyVmT:αsT, or PyVmT:Neo cells (Fig. 2,A, right). Doubling time was similar for all three lines, and their proliferation was unaffected by (added) 0.1 to 10 ng/mL TGF-β. Addition of TGF-β1 to all three cell lines induced phosphorylation of Smad2 (Fig. 2,B). In low serum, the PyVmT:αsT cells exhibited more than twofold the rate of apoptosis compared with controls, which was rescued by exogenous TGF-β1 (Fig. 2,C). In transwell assays, PyVmT:αsT cells migrated in response to 10% FCS at a similar rate as PyVmT or PyVmT:Neo cells. However, in response to 0.5% FCS, migration of PyVmT:αsT cells was reduced >50% compared with both control cells. Exogenous TGF-β1 rescued the impaired motility of PyVmT:αsT cells, and this rescue was blocked by Fc:TβRII (Fig. 2,D). Anchorage-independent growth of PyVmT cells produced an average of 460 colonies in soft agar compared with 59 PyVmT:αsT colonies. Finally, addition of TGF-β1 also rescued the impaired PyVmT:αsT colony growth in soft agar (Fig. 2 E).

PyVmT or PyVmT:αsT cells were implanted into mammary fat pads of FVB mice. Mice were monitored twice weekly by palpation to determine tumor latency. Whereas mice harboring PyVmT or PyVmT:Neo cells formed palpable tumors with an average latency of 35 days, the average latency was 53 days in mice injected with PyVmT:αsT cells (Fig. 3,A). PyVmT:αsT tumors were 6.2-fold smaller than PyVmT tumors when measured 100 days after initial tumor palpation (Fig. 3,B). PyVmT:αsT tumors exhibited a lower histologic grade and contained fewer mitotic figures, blood vessels, and regions of necrosis than control tumors. PyVmT:αsT tumors were histologically similar to PyVmT tumors in mice treated twice weekly with Fc:TβRII (Fig. 3,B). TUNEL analysis showed marked increased apoptosis in PyVmT:αsT tumors compared with PyVmT tumors and in control tumors treated with Fc:TβRII (Fig. 3,D). Tumor cell proliferation as measured by PCNA immunohistochemistry was similar in tumors of all three genotypes (Fig. 3 D).

To determine whether administration of Fc:TβRII modified immune mechanisms, we measured the ability of FVB mouse splenocytes to lyse ⁵¹Cr-labeled Yac-1 cells ex vivo. Cell lysis activity was similar in splenocytes from Fc:TβRII-treated and -untreated mice (data not shown). To avoid a primary tumor lead-time bias, lung metastases were examined 100 days following initial tumor palpation. PyVmT tumors produced 11-fold more surface lung metastases than PyVmT:αsT tumors. Inhibition of TGF-β in vivo with Fc:TβRII markedly reduced PyVmT lung metastases (Fig. 3 C). Collectively, these results suggest that at autocrine TGF-β, at least in part, enhances cancer progression by increasing cell survival, motility, and metastases.

A constitutively active simian TGF-β1 (TGF-β1^S223/225) transgene under the control of the TetOp₇ promoter was constructed. The TetOp₇ promoter contains seven tandem repeats of a sequence that is trans-activated by the transcription factor rtTA only in the presence of DOX (39). The TetOp-TGF-β1^S223/225 transgene was used to generate two founders, 1 and 35, of 56 pronuclear injections. Both founder animals could pass on the transgene to their offspring and were crossed with MMTV-rtTA mice to generate double transgenics (Fig. 4,A). Treatment of bitransgenic r/T with DOX induced expression of the TGF-β1 transgene product as determined by Northern analysis. This was not observed in MMTV-rtTA or TetOp-TGF-β1^S223/225 female mice treated with DOX. Bitransgenic r/T female mice not receiving DOX did not express simian TGF-β1 RNA (Fig. 4 B).

MMTV-rtTA female mice were bred with TetOp-TGF-β1^S223/225 male mice, and pregnancies were timed such that day 0.5 was the first morning postcoitus. Beginning at 7.5 days postcoitus and lasting through lactation, pregnant and nursing female mice were given DOX in the drinking water. After weaning, pups were given DOX in their own drinking water. In this manner, mice were maintained on DOX from 7.5 days postcoitus through 12 weeks of age (14 weeks total). Consistent with the reported inhibitory effect of TGF-β1 on mammary gland development, glands from DOX-treated r/T mice showed a severe delay in ductal progression compared with wild-type or single transgenic mice (Fig. 4,C). Expression of the active TGF-β1 was detected at low levels in mammary epithelium from untreated mice or from single transgenic mice treated with DOX, whereas higher levels of active TGF-β1 protein were detected in DOX-treated r/T epithelial cells (Fig. 4,D). Nuclear localization of Smad2 was detected in epithelial cell nuclei of r/T mice treated with DOX but remained cytoplasmic in single transgenic mice (Fig. 4 E).

MMTV-PyVmT transgenic mice develop mammary tumors with an average latency of 53 days and form lung metastases with 100% penetrance by 100 days of age. MMTV-PyVmT mice were crossed with r/T mice to generate triple transgenic MMTV-PyVmT-r/T (PrT) mice or double transgenic Pr or PT controls. In the absence of DOX, tumors arose with similar latencies in Pr, PT, and PrT mice (Fig. 5,A). Because the average tumor latency for each group was ∼8 weeks, DOX was administered no earlier than 9 weeks to allow tumor formation to occur unabated. Whole mounts of right inguinal (no. 4) mammary glands from 9-week-old PT, Pr, and PrT mice showed no differences in gland morphology (Fig. 5,B). Mice were treated with DOX from weeks 9 to 13 or 11 to 13. The majority of DOX-treated PrT mice exhibited signs of respiratory distress at 13 weeks. Active TGF-β1 was detected at week 13 only in mammary tumors from triple transgenic PrT mice treated for 1 or 4 weeks with DOX but not in tumors from untreated PrT mice or from bigenic Pr or PT mice (Fig. 5,C and data not shown). By immunofluorescence, Smad2 localized in tumor cell nuclei from DOX-treated PrT mice but not from control (−DOX) mice (Fig. 5,D). This also was evident in immunoblot analysis of nuclear fractions from DOX-induced and uninduced PrT tumors (Fig. 5,E). P-Smad2 was undetectable in the nuclear fractions (not shown). Serum levels of active TGF-β1 at week 13 were 2155 ± 483 pg/mL (n = 6) and 7400 ± 748 pg/mL (n = 6) in the untreated versus DOX-treated PrT mice, respectively. Finally, immunoblot analyses showed overall moderately higher levels of P-Akt and P-MAPK in tumor lysates from DOX-treated than DOX-untreated PrT mice and tumors from mice lacking the TetOp-TGF-β1^S223/225 transgene (Fig. 5 F).

Lung metastases were counted at 13 weeks of age. After 4 weeks of DOX treatment, PrT mice displayed an average of 162 ± 15.9 (n = 15) surface lung metastases compared with 17.6 ± 2.6 (n = 11) in Pr mice (Fig. 6,A) or 24.9 ± 3.9 in PT mice (not shown). Primary mammary tumors from DOX-treated PrT, Pr, or PT mice were histologically similar, and all three genotypes developed tumors in 10 of 10 mammary glands with no changes in tumor volume during DOX induction. To determine whether a shorter duration of TGF-β induction also increased lung metastases, mice were treated with DOX from weeks 11 to 13 (Fig. 6,C and D). Again, the number of lung metastases in PrT mice outnumbered those in Pr mice or PT mice by >10-fold, but they were fewer than in PrT mice treated with DOX for 4 weeks (compare Fig. 6,A and C), suggesting a time dependence for the effect of induced TGF-β1. The number of lung metastases in untreated PrT mice was not different from the number observed in untreated Pr or PT mice (Fig. 6 E and F).

Reduced apoptosis was observed in primary tumors and lung metastases in DOX-treated PrT mice at 13 weeks: 5.5 and 1.6% tumor cells were TUNEL positive in the untreated versus DOX-treated primary tumors (n = 4 each), respectively; 2.2 and 0.4% cells were TUNEL positive in the lung metastases from untreated versus DOX-treated mice (n = 4 each), respectively (Fig. 7,A). By PCNA immunohistochemistry, tumor cell proliferation was the same in untreated and DOX-treated tumors (not shown). Finally, addition of DOX induced a fibroblastoid morphology in PrT triple transgenic but not PyVmT primary cultures. These changes were similar to those induced by recombinant TGF-β1 in both cell types (Fig. 7 B), suggesting they were mediated by DOX-induced TGF-β in an autocrine fashion.

TGF-β exhibits a tumor suppressor and a tumor promoter role. In support of its role as a tumor suppressor, mice heterozygous for the Tgfb1 gene develop an increased number of carcinogen-induced liver and lung tumors (40). Mice with targeted disruption of Tgfb1 or Smad3 or heterozygous disruption of Smad4 develop tumors of the gastrointestinal tract (6, 11, 12). Conversely, tissue-specific overexpression of active TGF-β1 in mice has been shown to reduce carcinogen- and oncogene-induced mammary cancers (9). A recent report indicates that a T29 C polymorphism in the TGFB1 gene results in increased serum levels of TGF-β1 and is associated with a reduced risk of breast cancer in postmenopausal women (41). The ability of excess TGF-β to prevent cancer does not necessarily imply that an inhibitory effect should be expected against established cancers. Exogenous TGF-β has never been shown to inhibit an established neoplasm in vivo nor has the administration of a TGF-β inhibitor resulted in either spontaneous tumor development or the acceleration of an already established cancer.

Conversely, several reports support a causal association between an excess of endogenous or exogenous TGF-β and tumor progression. For example, overexpression of activated type I TGF-β receptor (29) or active TGF-β1 (30) under the control of a mammary-specific promoter recently was shown to accelerate metastases from neu-induced primary mammary tumors in transgenic mice. However, because in these studies TGF-β expression or signaling was not regulated in a controlled fashion, they do not address the role of TGF-β on late tumor progression, a scenario in which TGF-β inhibitors will be tested first in the near future. Thus, we developed a triple transgenic mouse model in which expression of active TGF-β1 in mammary tumors could be temporally controlled. We show herein that DOX-mediated induction of active TGF-β1 in late mammary tumors markedly accelerated metastases.

Several data suggest that this was caused, at least in part, by a direct effect of TGF-β on tumor cells. First, overexpression of detectable active TGF-β1 was largely limited to mammary epithelium also expressing PyVmT (Fig. 5). Second, Smad2 exhibited higher nuclear localization in tumor cells from DOX-treated mice. Of note, we were unable to detect high P-Smad2 by immunoblot in the same DOX-treated tumors in which Smad2 levels were higher in the nucleus (Fig. 5,F), suggesting the possibility that phosphorylation and nuclear localization might be temporally dissociable. Third, tumor lysates from DOX-treated mice contained overall higher levels of P-Akt and P-MAPK. Fourth, primary and metastatic tumor cells exhibited reduced apoptosis while in the presence of more than threefold higher serum levels of active TGF-β1. Fifth, addition of DOX to primary cultures of triple transgenic (PrT) cells induced fibroblastoid morphology consistent with epithelial-to-mesenchymal transition (Fig. 4). Although we cannot rule out paracrine effects of the induced TGF-β1 on stromal, immune, or endothelial cells, these data support the probable contribution of an autocrine effect on tumor cells to explain the enhanced metastatic progression. Determining the exact contribution of autocrine versus paracrine effect of TGF-β in vivo is difficult and will require examining the net effect in situ (or lack of) of TGF-β inhibitors in tumor and nontumor compartments in mice bearing TβRII-null tumors.

Additional support of at least a partial role for autocrine TGF-β in the progression of PyVmT tumors is provided by the tumor cells stably expressing antisense TGF-β1. These cells exhibited impaired basal motility, survival, and anchorage-independent growth in vitro and reduced tumorigenicity and metastases in vivo. Interestingly, expression of antisense TGF-β or administration of Fc:TβRII inhibits metastases. This correlation implies the possibility that the exogenous inhibitor was at least partially working by blocking autocrine/paracrine TGF-β inputs to the tumor cells. Of note, Fc:TβRII did not alter mouse splenocyte natural killer function, arguing against a general immunologic effect to explain its antitumor action. TGF-β2 was expressed at lower levels than TGF-β1 in PyVmT cells. TGF-β1 expression was inhibited by the antisense, whereas TGF-β2 expression remained intact. This suggests that TGF-β2 expression was unable to compensate for the antisense-mediated reduction in tumorigenicity. This speculation also is consistent with the inhibitory effect of Fc:TβRII on PyVmT cells and tumors (Figs. 1,D and E and 3 B–D). Fc:TβRII should not block TGF-β2 because of the low affinity of the type II receptor for TGF-β2 (42). Thus, its antimetastatic effect is likely caused by binding TGF-β1 and/or TGF-β3.

Blockade of endogenous TGF-β in parental and vector control cells with a soluble Fc:TβRII fusion protein reduced basal cell motility and survival and down-regulated basal levels of P-Akt and P-MAPK (Fig. 1), suggesting that these phenotypic and biochemical responses were at least in part regulated by autocrine TGF-β ligands. These results are consistent with the reported antitumor effect of TGF-β blockade in MMTV-PyVmT transgenic mice, in which treatment with sTβRII:Fc results in increased tumor cell apoptosis and inhibition of tumor cell P-Akt levels, motility, intravasation, and metastases (43). The reduction in P-Akt is somewhat surprising in that middle T per se can engage PI3K/Akt signaling (44). However, this result implies that some oncogenes can engage autocrine TGF-β on the induction of non-Smad signaling pathways; in turn, oncogene-transformed cells may become partially dependent on (permissive) TGF-β-induced signals for tumor progression.

The results presented herein have important clinical implications because they suggest that the induction of high systemic and/or tumor levels of TGF-β even for a short time can accelerate cancer progression. Elevated levels of plasma TGF-β are detected in patients with cancer, and in some cases, they predict for early metastatic recurrences (45, 46, 47, 48). Several anticancer therapies, many of which are ineffective against late tumors, can induce TGF-β systemically or in situ (49, 50, 51, 52, 53, 54, 55). These data coupled with the antiapoptotic effects of TGF-β on transformed cells (43, 56, 57) suggest the possibility that an excess of TGF-β could contribute to drug resistance For example, TGF-β1 and TGF-β3 have been shown to protect tumor cells from tumor necrosis factor and cell cycle-selective chemotherapeutics (58, 59). Moreover, tumors resistant to anticancer therapies overexpress TGF-βs (60, 61), and TGF-β blockade has been shown to reverse this resistance (62). Thus, we surmise that, in addition to directly facilitating the natural progression of established tumors, the (fortuitous) induction of TGF-β in a tumor host by variables that remain to be fully elucidated also may counteract the effects of anticancer therapies. The results presented in this article support the prospective investigation of these hypotheses.