Logo of pnasPNASInfo for AuthorsSubscriptionsAboutThis Article
Proc Natl Acad Sci U S A. 2003 Jul 22; 100(15): 8621–8623.
Published online 2003 Jul 14. doi:  10.1073/pnas.1633291100
PMCID: PMC166359

The two faces of transforming growth factor β in carcinogenesis

Hero or villain? Trustworthy guardian of normal homeostasis or double agent? Over the past two decades, the perceived role of transforming growth factor β (TGF-β) in carcinogenesis has undergone more plot twists than an Agatha Christie mystery. The initial experiments leading to the discovery of TGF-β and its naming as a “transforming” growth factor were based on its ability to induce malignant behavior of normal fibroblasts, leading to the notion that TGF-β might be a key factor in uncoupling a cell from normal growth control in such a way that it could become tumorigenic (1). However, at the time, this presumed function of the protein was difficult to reconcile with its ubiquitous pattern of expression in normal tissues, including its prevalence in human platelets. The next twist in the story came several years later, when it emerged that TGF-β has profound growth-suppressive effects on many cells, including epithelial cells and lymphoid cells, which form the basis of the majority of human cancers. At this point TGF-β began to be given serious consideration as a candidate tumor suppressor gene (2). Indeed, data from both experimental model systems and studies of human cancers clearly show that not only the ligand itself but also its downstream elements, including its receptors, and its primary cytoplasmic signal transducers, the Smad proteins, are important for suppressing primary tumorigenesis in many organs (3, 4).

While solid credentials were being established for the role of TGF-β as a good citizen in the battle to maintain cellular order, a darker side was emerging. It is now appreciated that metastasis of many different types of tumor cells actually requires TGF-β activity and that, in the context of advanced disease, it actually has prooncogenic effects (3, 5). To date, understanding of this complex, dual role of TGF-β in carcinogenesis has come principally from inference based on the synthesis of studies of the behavior of many different tumor cell lines and many different animal model systems (Fig. 1). But in a recent issue of PNAS Siegel et al. (6) convincingly demonstrated this duality in a single model of mammary cancer that is metastatic to lung, by using bitransgenic mice expressing forms of the Neu oncogene driven by the mouse mammary virus long-terminal repeat (MMTV) and either constitutively activated or dominant negative forms of the TGF-β receptors similarly directed to the mammary gland. This model system is highly relevant to human breast cancer as amplification of erbb2/her2/neu, encoding an epidermal growth factor receptor family tyrosine kinase, is found in >30% of human breast cancers (7) and changes in expression of both the TGF-β ligand and response system occur during cancer progression (810). The results of Siegel et al. (6) show that activation of TGF-β signaling delays the appearance of primary mammary tumors, whereas tumors appear earlier in mice in which TGF-β signaling has been compromised, consistent with a tumor suppressor role of the pathway in the earlier stages of tumorigenesis. In contrast, these same manipulations of TGF-β signaling have opposite effects on formation of spontaneous lung metastases in the mice. Mice expressing an activated TGF-β receptor exhibit an increased percentage of metastatic foci that have extravastated from the vasculature, consistent with a prooncogenic effect for TGF-β in late-stage disease.

Fig. 1.
TGF-β switches from tumor suppressor in the premalignant stages of tumorigenesis to prooncogene at later stages of disease leading to metastasis. Progression to metastatic disease is generally accompanied by decreased or altered TGF-β ...

This dual- or multifunctionality of TGF-β has plagued and confused researchers for years. Indeed, it seems that for every presumed function of TGF-β, an example can be found in which it displays exactly the opposite activity (11). In part, this stems from the fact that, unlike many peptide growth factors that act on a restricted set of target cells, TGF-β is produced by and can act on nearly every cell type, with the consequence that the output of its signaling pathway is both cell type- and context-dependent. Recent advances in understanding molecular aspects of the TGF-β receptor serine/threonine kinases and their interacting partners, as well as the signal transduction pathways and effects on the transcriptome, now offer the potential for understanding the molecular mechanisms contributing to these superficially opposite effects of TGF-β, including especially its transition from tumor suppressor to oncogene during tumorigenesis (3, 12).

The accepted model of the tumor suppressor role of TGF-β in carcinogenesis is that it is critical for maintaining homeostatic control of growth not only in premalignant cells but also in cells progressing through the early stages of carcinogenesis. Central to its growth inhibitory effects, at least in vitro, are its abilities to suppress expression of c-Myc and to enhance expression of the cyclin-dependent kinase inhibitors p15INK4b and p21CIP1 (13). Effects of TGF-β on genomic stability, replicative senescence, and apoptosis also come into play in particular cell types (14, 15). Although less well understood molecularly, effects of TGF-β on tissue architecture and tumor stroma are also important in tumorigenesis (16), and likely contribute to the outcome in human disease and in animal models such as the MMTV-Neu model used by Siegel et al. (6). Although these researchers did show reduced proliferation in primary mammary tumors in which TGF-β signaling had been activated, the contribution of other TGF-β-dependent mechanisms, such as induction of apoptosis, was not explored.

Many cancer researchers are now focused on attempts to understand and control metastasis, because metastases are responsible for most patient deaths. The idea that TGF-β exerts an opposite, prooncogenic role in metastatic disease, and possibly is even required for metastasis in some cells, is not a new one. Analysis of immunohistochemical staining of TGF-β1 in breast cancers showed enhanced expression in lymph node metastases with preferential expression at the advancing edges of the tumors (10). Consistent with this observation, treatment of metastatic breast cancer cell lines with TGF-β in vitro enhanced metastases (17), whereas suppression of TGF-β receptor function markedly reduced their ability to metastasize (18). Indeed, TGF-β may even play a specific role in directing metastatic cells to particular organ sites such as bone, which is a common site of metastatic foci of breast and prostate cancer. For example, studies have shown that TGF-β/Smad and p38 signaling pathways cooperate to promote metastasis of human breast cancer cells to bone by inducing expression of the osteolytic factor, PTH-related protein (PTHrP) (19, 20). Similar to the approach that Siegel et al. (6) have taken to analyze effects of TGF-β on tumorigenesis in MMTV-Neu mice, these researchers showed that expression of a constitutively active TGF-β type I receptor in the breast cancer cells enhanced expression of PTHrP and occurrence of osteolytic bone metastases, whereas expression of a dominant negative TGF-β type II receptor had the opposite effects (19, 20). Together, these studies clearly point to tumor cell autonomous oncogenic effects of TGF-β and its gene targets on metastases.

The mechanisms whereby metastatic tumor cells break off from their primary site and invade the vasculature, travel through the blood stream to a remote site, and eventually colonize and grow in that organ are still being debated (21). Indeed, there is some suggestion that the mechanisms may differ between organ sites or tumor cell types. One theory put forth by Chambers and colleagues (21, 22) suggests that extravasation into the target organ is an efficient step for most cells and that the rate-limiting steps for metastasis are the successful breaking of dormancy to establish micrometastases in the new environment, and the subsequent transition to vascularized macrometastases. In an alternative view, Al-Mehdi, Muschel, and colleagues (23, 24) suggest that for some tumor cells, hematogenous metastases to lung originate from adherence of tumor cells to the vasculature of the target tissue and that the cells proliferate initially within the vessel, before and without the need for extravasation. In this model, the rate-limiting step is escape from apoptosis and, eventually, extravasation after intravascular expansion. Of interest here are the observations of Siegel et al. (6), which show that whereas the total number of lung metastases in bigenic MMTV-Neu-activated TβRI mice did not differ from that found in monogenic MMTV-Neu mice, activation of TGF-β signaling in the tumor cells enhanced the proportion of extravasated metastases. Surprisingly, tumor cells in which signaling was suppressed by expression of a dominant negative receptor showed no difference in either the number of metastatic foci or the proportion of metastases that had extravasated. In contrast, systemic expression of a soluble TGF-β antagonist has been shown to decrease the incidence of metastatic disease in the MMTV-Neu model (25). Together, the two approaches suggest that although the extravasation step may require cell-autonomous action of TGF-β on the tumor cell, as implied by earlier studies (17), there are likely to be additional rate-limiting steps that involve other aspects of metastasis contributed by host stromal cells (Fig. 1).

TGF-β may play a role in directing metastatic cells to particular organ sites.

To take these observations beyond the phenomenological and enable development of new therapeutic approaches, we must now address the mechanisms whereby the TGF-β signaling pathway and/or its gene targets change as cells acquire the capacity to invade and metastasize. It is known that as cells progress toward fully malignant tumor cells, they undergo changes that result in reduced expression of TGF-β receptors, increased expression of TGF-β ligands, and resistance to inhibition of growth by TGF-β (3). Use of small molecule inhibitors of the TGF-β type I receptor kinase has shown that inhibition of growth requires a sustained signal of at least 12–14 h, but that induction of certain immediate early target genes of TGF-β requires only a transient signal (26, 27). Thus, we may find that as signaling flux is reduced in more malignant cells, the set of gene targets of TGF-β will become restricted to those that require a lower signaling intensity or a signal of shorter duration. This can now be approached in systems such as that described by Siegel et al. (6) by gene profiling and analysis of mechanistic correlates.

Another question to be addressed is whether TGF-β utilizes the same or different signaling pathways to mediate its tumor suppressor and oncogenic effects. The predominant pathway for signal transduction from the TGF-β receptor serine/threonine kinases is mediated by cytoplasmic proteins called Smads. In this pathway, Smad2 and Smad3 are phosphorylated directly by the TGF-β type I receptor kinase and after partnering with the common mediator, Smad4, translocate to the nucleus, where they regulate transcription of target genes (12). Although tumor suppressor activity has been attributed to all of the components of this pathway, including TGF-β itself, only in rather rare instances do the genes follow the classic Knudsonian paradigm for a tumor suppressor. For example, the TGF-β ligand shows haplo-insufficiency as a tumor suppressor and the second allele is not deleted in tumorigenesis, but rather is overexpressed as cells acquire more malignant properties (28). Receptors are often transcriptionally repressed in such a way that some function still remains (12, 29). As evidence of the dual role of the TGF-β receptor in tumorigenesis, the prognosis of patients with hereditary nonpolyposis colorectal cancer tumors, where receptors are biallelically lost by microsatellite instability, is more favorable that that of patients with colon cancers, where receptor expression is merely repressed (30). This pattern is reiterated at the level of the Smad proteins, which are mutationally inactivated in only a small subset of cancers, but for which there exists a large variety of epigenetic mechanisms to suppress their activity in carcinogenesis (4). Therefore, the possibility that these mechanisms have evolved to permit the tumor cell to reactivate this pathway as needed in metastasis must be considered. TGF-β also activates mitogen-activated protein kinase pathways, and it has been suggested that a possible imbalance between signaling through these pathways and the Smad pathway or cooperation between these pathways might contribute to the prooncogenic role of TGF-β (3, 31, 32).

We have recently developed a system to examine the dual roles of TGF-β as tumor suppressor and oncogene by using a series of breast cancer cell lines each derived from Ha-ras transformed MCF10A cells, and thus sharing a common origin, but exhibiting distinct features ranging from premalignant to fully invasive and metastatic (33). By manipulating either TGF-β receptor function (B. Tang, M. Vu, T. Booker, S. J. Santner, F. R. Miller, M. R. Anver, and L.M.W., unpublished data) or the activity of the Smad2 and/or Smad3 pathway in these cells (F. Tian, S. Byfield, W. T. Parks, S. Yoo, A. Felici, B., Tang, E. Piek, L.M.W., and A.B.R., unpublished data), we have been able to show that either reduction in TGF-β receptor function or inhibition of the Smad2/3 pathway, as might be found in a tumor cell, enhance the tumorigenesis of xenografted premalignant and well differentiated carcinoma cell lines, but substantially reduce the formation of metastatic foci in the lung after tail vein injection of a metastatic cell line of common origin. Thus, our results demonstrating the dual tumor suppressor/oncogene role of TGF-β in this genetically and biochemically tractable system parallel those of Siegel et al. (6). Moreover, by using this approach, we have been able to demonstrate that the same signaling pathway mediated by Smad2 and/or Smad3 is responsible for both of these activities of TGF-β.

Ultimately, it is hoped that the use of either a mouse model system, such as that described by Siegel et al. (6), or a cell-based system, such as that described above, will provide information about the specific pathways and gene targets that effect this unique switch from tumor suppressor to oncogene. This information is critical for design of new approaches to prevent metastasis as well as for understanding effects of inhibitors of this pathway that might be used to treat other chronic diseases, including fibrotic diseases (26). It is very promising in this regard that two recent studies have shown that an anti-ligand approach based on use of a soluble TGF-β-receptor selectively inhibits metastatic disease in the absence of effects either on normal physiology or on tumorigenesis at the primary site (25, 34). This exciting result adds yet another level of complexity to this TGF-β system by suggesting that the form or accessibility of the ligand is different in primary and metastatic sites. Although TGF-β still remains elusive in terms of our understanding of its multifunctional modes of action, we are moving ever closer to unraveling this mystery at a molecular level and to design of new therapeutic approaches directed toward modulating its activities.


See companion article on page 8430 in issue 14 of volume 100.


1. Roberts, A. B. & Sporn, M. B. (1985) Cancer Surv. 4, 683–705. [PubMed]
2. Wakefield, L. M. & Sporn, M. B. (1990) Immunol. Ser. 51, 217–243. [PubMed]
3. Wakefield, L. M. & Roberts, A. B. (2002) Curr. Opin. Genet. Dev. 12, 22–29. [PubMed]
4. de Caestecker, M. P., Piek, E. & Roberts, A. B. (2000) J. Natl. Cancer Inst. 92, 1388–1402. [PubMed]
5. Akhurst, R. J. & Derynck, R. (2001) Trends Cell Biol. 11, S44–S51. [PubMed]
6. Siegel, P. M., Shu, W., Cardiff, R. D., Muller, W. J. & Massagué, J. (2003) Proc. Natl. Acad. Sci. USA 100, 8430–8435. [PMC free article] [PubMed]
7. Slamon, D. J., Clark, G. M., Wong, S. G., Levin, W. J., Ullrich, A. & McGuire, W. L. (1987) Science 235, 177–182. [PubMed]
8. Gobbi, H., Arteaga, C. L., Jensen, R. A., Simpson, J. F., Dupont, W. D., Olson, S. J., Schuyler, P. A., Plummer, W. D. & Page, D. L. (2000) Histopathology 36, 168–177. [PubMed]
9. Gobbi, H., Dupont, W. D., Simpson, J. F., Plummer, W. D. J., Schuyler, P. A., Olson, S. J., Arteaga, C. L. & Page, D. L. (1999) J. Natl. Cancer Inst. 91, 2096–2101. [PubMed]
10. Dalal, B. I., Keown, P. A. & Greenberg, A. H. (1993) Am. J. Pathol. 143, 381–389. [PMC free article] [PubMed]
11. Sporn, M. B. & Roberts, A. B. (1990) Cell Regul. 1, 875–882. [PMC free article] [PubMed]
12. Massagué, J., Blain, S. W. & Lo, R. S. (2000) Cell 103, 295–309. [PubMed]
13. Chen, C. R., Kang, Y. & Massagué, J. (2001) Proc. Natl. Acad. Sci. USA 98, 992–999. [PMC free article] [PubMed]
14. Glick, A. B., Weinberg, W. C., Wu, I. H., Quan, W. & Yuspa, S. H. (1996) Cancer Res. 56, 3645–3650. [PubMed]
15. Stampfer, M. R., Garbe, J., Levine, G., Lichtsteiner, S., Vasserot, A. P. & Yaswen, P. (2001) Proc. Natl. Acad. Sci. USA 98, 4498–4503. [PMC free article] [PubMed]
16. Engle, S. J., Hoying, J. B., Boivin, G. P., Ormsby, I., Gartside, P. S. & Doetschman, T. (1999) Cancer Res. 59, 3379–3386. [PubMed]
17. Welch, D. R., Fabra, A. & Nakajima, M. (1990) Proc. Natl. Acad. Sci. USA 87, 7678–7682. [PMC free article] [PubMed]
18. McEarchern, J. A., Kobie, J. J., Mack, V., Wu, R. S., Meade-Tollin, L., Arteaga, C. L., Dumont, N., Besselsen, D., Seftor, E., Hendrix, M. J., et al. (2001) Int. J. Cancer 91, 76–82. [PubMed]
19. Kakonen, S. M., Selander, K. S., Chirgwin, J. M., Yin, J. J., Burns, S., Rankin, W. A., Grubbs, B. G., Dallas, M., Cui, Y. & Guise, T. A. (2002) J. Biol. Chem. 277, 24571–24578. [PubMed]
20. Yin, J. J., Selander, K., Chirgwin, J. M., Dallas, M., Grubbs, B. G., Wieser, R., Massagué, J., Mundy, G. R. & Guise, T. A. (1999) J. Clin. Invest. 103, 197–206. [PMC free article] [PubMed]
21. Chambers, A. F., Groom, A. C. & MacDonald, I. C. (2002) Nat. Rev. Cancer 2, 563–572. [PubMed]
22. MacDonald, I. C., Groom, A. C. & Chambers, A. F. (2002) BioEssays 24, 885–893. [PubMed]
23. Wong, C. W., Song, C., Grimes, M. M., Fu, W., Dewhirst, M. W., Muschel, R. J. & Al Mehdi, A. B. (2002) Am. J. Pathol. 161, 749–753. [PMC free article] [PubMed]
24. Al Mehdi, A. B., Tozawa, K., Fisher, A. B., Shientag, L., Lee, A. & Muschel, R. J. (2000) Nat. Med. 6, 100–102. [PubMed]
25. Yang, Y. A., Dukhanina, O., Tang, B., Mamura, M., Letterio, J. J., MacGregor, J., Patel, S. C., Khozin, S., Liu, Z. Y., Green, J., et al. (2002) J. Clin. Invest. 109, 1607–1615. [PMC free article] [PubMed]
26. Inman, G. J., Nicolas, F. J., Callahan, J. F., Harling, J. D., Gaster, L. M., Reith, A. D., Laping, N. J. & Hill, C. S. (2002) Mol. Pharmacol. 62, 65–74. [PubMed]
27. Randall, R. A., Germain, S., Inman, G. J., Bates, P. A. & Hill, C. S. (2002) EMBO J. 21, 145–156. [PMC free article] [PubMed]
28. Tang, B., Bottinger, E. P., Jakowlew, S. B., Bagnall, K. M., Mariano, J., Anver, M. R., Letterio, J. J. & Wakefield, L. M. (1998) Nat. Med. 4, 802–807. [PubMed]
29. Kim, S. J., Im, Y. H., Markowitz, S. D. & Bang, Y. J. (2000) Cytokine Growth Factor Rev. 11, 159–168. [PubMed]
30. Watanabe, T., Wu, T. T., Catalano, P. J., Ueki, T., Satriano, R., Haller, D. G., Benson, A. B., III, & Hamilton, S. R. (2001) N. Engl. J. Med. 344, 1196–1206. [PMC free article] [PubMed]
31. Janda, E., Lehmann, K., Killisch, I., Jechlinger, M., Herzig, M., Downward, J., Beug, H. & Grunert, S. (2002) J. Cell Biol. 156, 299–314. [PMC free article] [PubMed]
32. Oft, M., Akhurst, R. J. & Balmain, A. (2002) Nat. Cell Biol. 4, 487–494. [PubMed]
33. Santner, S. J., Dawson, P. J., Tait, L., Soule, H. D., Eliason, J., Mohamed, A. N., Wolman, S. R., Heppner, G. H. & Miller, F. R. (2001) Breast Cancer Res. Treat. 65, 101–110. [PubMed]
34. Muraoka, R. S., Dumont, N., Ritter, C. A., Dugger, T. C., Brantley, D. M., Chen, J., Easterly, E., Roebuck, L. R., Ryan, S., Gotwals, P. J., et al. (2002) J. Clin. Invest. 109, 1551–1559. [PMC free article] [PubMed]

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • Cited in Books
    Cited in Books
    PubMed Central articles cited in books
  • Compound
    PubChem Compound links
  • PubMed
    PubMed citations for these articles
  • Substance
    PubChem Substance links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...